| Title: | Fast Spatial and Spatio-Temporal Regression using Moran Eigenvectors |
|---|---|
| Description: | A collection of functions for estimating spatial and spatio-temporal regression models. Moran eigenvectors are used as spatial basis functions to efficiently approximate spatially dependent Gaussian processes (i.e., random effects eigenvector spatial filtering; see Murakami and Griffith 2015 <doi: 10.1007/s10109-015-0213-7>). The implemented models include linear regression with residual spatial dependence, spatially/spatio-temporally varying coefficient models (Murakami et al., 2017, 2024; <doi:10.1016/j.spasta.2016.12.001>,<doi:10.48550/arXiv.2410.07229>), spatially filtered unconditional quantile regression (Murakami and Seya, 2019 <doi:10.1002/env.2556>), Gaussian and non-Gaussian spatial mixed models through compositionally-warping (Murakami et al. 2021, <doi:10.1016/j.spasta.2021.100520>). |
| Authors: | Daisuke Murakami [aut, cre] |
| Maintainer: | Daisuke Murakami <[email protected]> |
| License: | GPL (>= 2) |
| Version: | 0.3.1 |
| Built: | 2024-11-12 03:25:44 UTC |
| Source: | https://github.com/cran/spmoran |
This function performs an additional learning of local variations in spatially varying coefficients (SVC). While the SVC model implemented in resf_vc or besf_vc can be less accurate for large samples (e.g., n > 5,000) due to a degeneracy/over-smoothing problem, this additional learning mitigates this problem by synthesizing/averaging the model with local SVC models. The resulting spatial prediction implemented in this function is expected to be more accurate than the resf_vc function.
Note that this function is not yet supported for spatio-temporal models with !is.null(meig$coords_z).
addlearn_local( mod, meig0 = NULL, x0 = NULL, xconst0=NULL, xgroup0=NULL, cl_num=NULL, cl=NULL, parallel=FALSE, ncores=NULL )addlearn_local( mod, meig0 = NULL, x0 = NULL, xconst0=NULL, xgroup0=NULL, cl_num=NULL, cl=NULL, parallel=FALSE, ncores=NULL )
mod |
|
meig0 |
Moran eigenvectors at prediction sites. Output from |
x0 |
Matrix of explanatory variables at prediction sites whose coefficients are allowed to vary across geographical space (N_0 x K). Default is NULL |
xconst0 |
Matrix of explanatory variables at prediction sites whose coefficients are assumed constant (or NVC) across space (N_0 x K_const). Default is NULL |
xgroup0 |
Matrix of group indeces at prediction sites that may be group IDs (integers) or group names (N_0 x K_g). Default is NULL |
cl_num |
Number of local sub-models being aggregated/averaged. If NULL, the number is determined so that the number of samples per sub-model equals approximately 600. Default is NULL |
cl |
Vector of cluster ID for each sample (N x 1). If specified, the local sub-models are given by this ID. If NULL, k-means clustering based on spatial coordinates is performed to obtain spatial clusters each of which contain approximately 600 samples. Default is NULL |
parallel |
If TRUE, the model is estimated through parallel computation. The default is FALSE if mod is an |
ncores |
Number of cores used for the parallel computation. If ncores = NULL and parallel = TRUE, the number of available cores - 2 is used. Default is NULL |
b_vc |
Matrix of estimated spatially varying coefficients (SVCs) on x (N x K) |
bse_vc |
Matrix of standard errors for the SVCs on x (N x k) |
z_vc |
Matrix of z-values for the SVCs on x (N x K) |
p_vc |
Matrix of p-values for the SVCs on x (N x K) |
c |
Matrix with columns for the estimated coefficients on xconst, their standard errors, z-values, and p-values (K_c x 4) |
b_g |
List of K_g matrices with columns for the estimated group effects, their standard deviations, and t-values |
s |
List of 2 elements summarizing variance parameters characterizing SVCs of each local sub-model. The first element contains standard deviations of each SVCs while the second elementcontains their Moran's I values that are scaled to take a value between 0 (no spatial dependence) and 1 (strongest positive spatial dependence). Based on Griffith (2003), the scaled Moran'I value is interpretable as follows: 0.25-0.50:weak; 0.50-0.70:moderate; 0.70-0.90:strong; 0.90-1.00:marked |
s_global |
The same variance parameters for the globa sub-model |
s_g |
Vector of standard deviations of the group effects |
e |
Error statistics. It includes residual standard error (resid_SE), adjusted conditional R2 (adjR2(cond)), restricted log-likelihood (rlogLik), Akaike information criterion (AIC), and Bayesian information criterion (BIC) |
pred |
Matrix of predicted values for y (pred) and their standard errors (pred_se) (N x 2) |
resid |
Vector of residuals (N x 1) |
cl |
Vector of cluster ID being used (N x 1) |
pred0 |
Matrix of predicted values for y (pred) and their standard errors (pred_se) at prediction sites (N_0 x 2) |
b_vc0 |
Matrix of estimated spatially varying coefficients (SVCs) at prediction sites (N_0 x K) |
bse_vc0 |
Matrix of standard errors for the SVCs at prediction sites (N_0 x k) |
z_vc0 |
Matrix of z-values for the SVCs at prediction sites (N x K) |
p_vc0 |
Matrix of p-values for the SVCs at prediction sites (N x K) |
other |
List of other outputs, which are internally used |
Daisuke Murakami
Murakami, D., Sugasawa, S., T., Seya, H., and Griffith, D.A. (2024) Sub-model aggregation-based scalable eigenvector spatial filtering: application to spatially varying coefficient modeling. Geographical Analysis, DOI: 10.1111/gean.12393.
require(spdep) data(house) dat0 <- data.frame(house@coords,house@data) dat <- dat0[dat0$yrbuilt>=1980,] ###### purpose 1: improve SVC modeling accuracy ###### ###### (i.e., addressing the over-smoothing problem) # y <- log(dat[,"price"]) x <- dat[,c("age","rooms")] xconst <- dat[,c("lotsize","s1994","s1995","s1996","s1997","s1998")] coords <- dat[ ,c("long","lat")] meig <- meigen_f( coords ) ## Not run # res0 <- resf_vc(y = y,x = x, xconst = xconst, meig = meig) # res <- addlearn_local(res0) # It adjusts SVCs to model local patterns # res ####### parallel version for very large samples (e.g., n >100,000) # bes0 <- besf_vc(y = y,x = x, xconst = xconst, coords=coords) # bes <- addlearn_local( bes0 ) ####### purpose 2: improve predictive accuracy ######## #samp <- sample( dim( dat )[ 1 ], 2500) #d <- dat[ samp, ] ## Data at observed sites #y <- log(d[,"price"]) #x <- d[,c("age","rooms")] #xconst <- d[,c("lotsize","s1994","s1995","s1996","s1997","s1998")] #coords <- d[ ,c("long","lat")] #d0 <- dat[-samp, ] ## Data at observed sites #y0 <- log(d0[,"price"]) #x0 <- d0[,c("age","rooms")] #xconst0 <- d0[,c("lotsize","s1994","s1995","s1996","s1997","s1998")] #coords0 <- d0[ ,c("long","lat")] #meig <- meigen_f( coords ) #res0 <- resf_vc(y = y,x = x, xconst = xconst, meig = meig) #meig0 <- meigen0( meig=meig, coords0=coords0 ) #res <- addlearn_local(res0, meig0=meig0, x0=x0, xconst0=xconst0) # #pred <- res$pred0 ## Predictive valuesrequire(spdep) data(house) dat0 <- data.frame(house@coords,house@data) dat <- dat0[dat0$yrbuilt>=1980,] ###### purpose 1: improve SVC modeling accuracy ###### ###### (i.e., addressing the over-smoothing problem) # y <- log(dat[,"price"]) x <- dat[,c("age","rooms")] xconst <- dat[,c("lotsize","s1994","s1995","s1996","s1997","s1998")] coords <- dat[ ,c("long","lat")] meig <- meigen_f( coords ) ## Not run # res0 <- resf_vc(y = y,x = x, xconst = xconst, meig = meig) # res <- addlearn_local(res0) # It adjusts SVCs to model local patterns # res ####### parallel version for very large samples (e.g., n >100,000) # bes0 <- besf_vc(y = y,x = x, xconst = xconst, coords=coords) # bes <- addlearn_local( bes0 ) ####### purpose 2: improve predictive accuracy ######## #samp <- sample( dim( dat )[ 1 ], 2500) #d <- dat[ samp, ] ## Data at observed sites #y <- log(d[,"price"]) #x <- d[,c("age","rooms")] #xconst <- d[,c("lotsize","s1994","s1995","s1996","s1997","s1998")] #coords <- d[ ,c("long","lat")] #d0 <- dat[-samp, ] ## Data at observed sites #y0 <- log(d0[,"price"]) #x0 <- d0[,c("age","rooms")] #xconst0 <- d0[,c("lotsize","s1994","s1995","s1996","s1997","s1998")] #coords0 <- d0[ ,c("long","lat")] #meig <- meigen_f( coords ) #res0 <- resf_vc(y = y,x = x, xconst = xconst, meig = meig) #meig0 <- meigen0( meig=meig, coords0=coords0 ) #res <- addlearn_local(res0, meig0=meig0, x0=x0, xconst0=xconst0) # #pred <- res$pred0 ## Predictive values
Parallel and memory-free implementation of RE-ESF-based spatial regression for very large samples. This model estimates residual spatial dependence, constant coefficients, and non-spatially varying coefficients (NVC; coefficients varying with respect to explanatory variable value).
besf( y, x = NULL, nvc = FALSE, nvc_sel = TRUE, coords, s_id = NULL, covmodel="exp", enum = 200, method = "reml", penalty = "bic", nvc_num = 5, maxiter = 30, bsize = 4000, ncores = NULL )besf( y, x = NULL, nvc = FALSE, nvc_sel = TRUE, coords, s_id = NULL, covmodel="exp", enum = 200, method = "reml", penalty = "bic", nvc_num = 5, maxiter = 30, bsize = 4000, ncores = NULL )
y |
Vector of explained variables (N x 1) |
x |
Matrix of explanatory variables (N x K) |
nvc |
If TRUE, NVCs are assumed on x. Otherwise, constant coefficients are assumed. Default is FALSE |
nvc_sel |
If TRUE, type of coefficients (NVC or constant) is selected through a BIC (default) or AIC minimization. If FALSE, NVCs are assumed across x. Alternatively, nvc_sel can be given by column number(s) of x. For example, if nvc_sel = 2, the coefficient on the second explanatory variable in x is NVC and the other coefficients are constants. The Default is TRUE |
coords |
Matrix of spatial point coordinates (N x 2) |
s_id |
Optional. ID specifying groups modeling spatially dependent process (N x 1). If it is specified, group-level spatial process is estimated. It is useful. e.g., for multilevel modeling (s_id is given by the group ID) and panel data modeling (s_id is given by individual location id). Default is NULL |
covmodel |
Type of kernel to model spatial dependence. The currently available options are "exp" for the exponential kernel, "gau" for the Gaussian kernel, and "sph" for the spherical kernel |
enum |
Number of Moran eigenvectors to be used for spatial process modeling (scalar). Default is 200 |
method |
Estimation method. Restricted maximum likelihood method ("reml") and maximum likelihood method ("ml") are available. Default is "reml" |
penalty |
Penalty to select type of coefficients (NVC or constant) to stablize the estimates. The current options are "bic" for the Baysian information criterion-type penalty (N x log(K)) and "aic" for the Akaike information criterion (2K) (see Muller et al., 2013). Default is "bic" |
nvc_num |
Number of basis functions used to model NVC. An intercept and nvc_num natural spline basis functions are used to model each NVC. Default is 5 |
maxiter |
Maximum number of iterations. Default is 30 |
bsize |
Block/badge size. bsize x bsize elements are iteratively processed during the parallelized computation. Default is 4000 |
ncores |
Number of cores used for the parallel computation. If ncores = NULL, the number of available cores - 2 is detected and used. Default is NULL |
b |
Matrix with columns for the estimated coefficients on x, their standard errors, z-values, and p-values (K x 4). Effective if nvc =FALSE |
c_vc |
Matrix of estimated NVCs on x (N x K). Effective if nvc =TRUE |
cse_vc |
Matrix of standard errors for the NVCs on x (N x K). Effective if nvc =TRUE |
ct_vc |
Matrix of t-values for the NVCs on x (N x K). Effective if nvc =TRUE |
cp_vc |
Matrix of p-values for the NVCs on x (N x K). Effective if nvc =TRUE |
s |
Vector of estimated variance parameters (2 x 1). The first and the second elements denote the standard deviation and the Moran's I value of the estimated spatially dependent component, respectively. The Moran's I value is scaled to take a value between 0 (no spatial dependence) and 1 (the maximum possible spatial dependence). Based on Griffith (2003), the scaled Moran'I value is interpretable as follows: 0.25-0.50:weak; 0.50-0.70:moderate; 0.70-0.90:strong; 0.90-1.00:marked |
e |
Vector whose elements are residual standard error (resid_SE), adjusted conditional R2 (adjR2(cond)), restricted log-likelihood (rlogLik), Akaike information criterion (AIC), and Bayesian information criterion (BIC). When method = "ml", restricted log-likelihood (rlogLik) is replaced with log-likelihood (logLik) |
vc |
List indicating whether NVC are removed or not during the BIC/AIC minimization. 1 indicates not removed whreas 0 indicates removed |
r |
Vector of estimated random coefficients on Moran's eigenvectors (L x 1) |
sf |
Vector of estimated spatial dependent component (N x 1) |
pred |
Vector of predicted values (N x 1) |
resid |
Vector of residuals (N x 1) |
other |
List of other outputs, which are internally used |
Daisuke Murakami
Griffith, D. A. (2003). Spatial autocorrelation and spatial filtering: gaining understanding through theory and scientific visualization. Springer Science & Business Media.
Murakami, D. and Griffith, D.A. (2015) Random effects specifications in eigenvector spatial filtering: a simulation study. Journal of Geographical Systems, 17 (4), 311-331.
Murakami, D. and Griffith, D.A. (2019) A memory-free spatial additive mixed modeling for big spatial data. Japan Journal of Statistics and Data Science. DOI:10.1007/s42081-019-00063-x.
require(spdep) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] xgroup <- boston.c[,"TOWN"] coords <- boston.c[,c("LON", "LAT")] ######## Regression considering spatially dependent residuals #res <- besf(y = y, x = x, coords=coords) #res ######## Regression considering spatially dependent residuals and NVC ######## (coefficients or NVC is selected) #res2 <- besf(y = y, x = x, coords=coords, nvc = TRUE) ######## Regression considering spatially dependent residuals and NVC ######## (all the coefficients are NVCs) #res3 <- besf(y = y, x = x, coords=coords, nvc = TRUE, nvc_sel=FALSE)require(spdep) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] xgroup <- boston.c[,"TOWN"] coords <- boston.c[,c("LON", "LAT")] ######## Regression considering spatially dependent residuals #res <- besf(y = y, x = x, coords=coords) #res ######## Regression considering spatially dependent residuals and NVC ######## (coefficients or NVC is selected) #res2 <- besf(y = y, x = x, coords=coords, nvc = TRUE) ######## Regression considering spatially dependent residuals and NVC ######## (all the coefficients are NVCs) #res3 <- besf(y = y, x = x, coords=coords, nvc = TRUE, nvc_sel=FALSE)
Parallel and memory-free implementation of SNVC modeling for very large samples. The model estimates residual spatial dependence, constant coefficients, spatially varying coefficients (SVCs), non-spatially varying coefficients (NVC; coefficients varying with respect to explanatory variable value), and SNVC (= SVC + NVC). Type of coefficients can be selected through BIC/AIC minimization. By default, it estimates a SVC model. SNVCs can be mapped just like SVCs. Unlike SVC models, SNVC model is robust against spurious correlation (multicollinearity), so, stable (see Murakami and Griffith, 2020). This function is not yet supported for spatio-temporal modeling.
Note: The SVC model can be less accurate for large samples due to a degeneracy/over-smoothing problem (see Murakami et al., 2023). The addlearn_local is useful to mitigate this problem (See the coding example below).
besf_vc( y, x, xconst = NULL, coords, s_id = NULL, x_nvc = FALSE, xconst_nvc = FALSE, x_sel = TRUE, x_nvc_sel = TRUE, xconst_nvc_sel = TRUE, nvc_num=5, method = "reml", penalty = "bic", maxiter = 30, tol = 1e-30, covmodel="exp",enum = 200, bsize = 4000, ncores=NULL )besf_vc( y, x, xconst = NULL, coords, s_id = NULL, x_nvc = FALSE, xconst_nvc = FALSE, x_sel = TRUE, x_nvc_sel = TRUE, xconst_nvc_sel = TRUE, nvc_num=5, method = "reml", penalty = "bic", maxiter = 30, tol = 1e-30, covmodel="exp",enum = 200, bsize = 4000, ncores=NULL )
y |
Vector of explained variables (N x 1) |
x |
Matrix of explanatory variables with spatially varying coefficients (SVC) (N x K) |
xconst |
Matrix of explanatory variables with constant coefficients (N x K_c). Default is NULL |
coords |
Matrix of spatial point coordinates (N x 2) |
s_id |
Optional. ID specifying groups modeling spatially dependent process (N x 1). If it is specified, group-level spatial process is estimated. It is useful for multilevel modeling (s_id is given by the group ID) and panel data modeling (s_id is given by individual location id). Default is NULL |
x_nvc |
If TRUE, SNVCs are assumed on x. Otherwise, SVCs are assumed. Default is FALSE |
xconst_nvc |
If TRUE, NVCs are assumed on xconst. Otherwise, constant coefficients are assumed. Default is FALSE |
x_sel |
If TRUE, type of coefficient (SVC or constant) on x is selected through a BIC (default) or AIC minimization. If FALSE, SVCs are assumed across x. Alternatively, x_sel can be given by column number(s) of x. For example, if x_sel = 2, the coefficient on the second explanatory variable in x is SVC and the other coefficients are constants. The Default is TRUE |
x_nvc_sel |
If TRUE, type of coefficient (NVC or constant) on x is selected through the BIC (default) or AIC minimization. If FALSE, NVCs are assumed across x. Alternatively, x_nvc_sel can be given by column number(s) of x. For example, if x_nvc_sel = 2, the coefficient on the second explanatory variable in x is NVC and the other coefficients are constants. The Default is TRUE |
xconst_nvc_sel |
If TRUE, type of coefficient (NVC or constant) on xconst is selected through the BIC (default) or AIC minimization. If FALSE, NVCs are assumed across xconst. Alternatively, xconst_nvc_sel can be given by column number(s) of xconst. For example, if xconst_nvc_sel = 2, the coefficient on the second explanatory variable in xconst is NVC and the other coefficients are constants.The Default is TRUE |
nvc_num |
Number of basis functions used to model NVC. An intercept and nvc_num natural spline basis functions are used to model each NVC. Default is 5 |
method |
Estimation method. Restricted maximum likelihood method ("reml") and maximum likelihood method ("ml") are available. Default is "reml" |
penalty |
Penalty to select type of coefficients (SNVC, SVC, NVC, or constant) to stablize the estimates. The current options are "bic" for the Baysian information criterion-type penalty (N x log(K)) and "aic" for the Akaike information criterion (2K) (see Muller et al., 2013). Default is "bic" |
maxiter |
Maximum number of iterations. Default is 30 |
tol |
The tolerance for matrix inversion. Some errors regarding singular fit can be avoided by reducing the value, but the output can be unstable. Default is 1e-30 |
covmodel |
Type of kernel to model spatial dependence. The currently available options are "exp" for the exponential kernel, "gau" for the Gaussian kernel, and "sph" for the spherical kernel |
enum |
Number of Moran eigenvectors to be used for spatial process modeling (scalar). Default is 200 |
bsize |
Block/badge size. bsize x bsize elements are iteratively processed during the parallelized computation. Default is 4000 |
ncores |
Number of cores used for the parallel computation. If ncores = NULL, the number of available cores - 2 is detected and used. Default is NULL |
b_vc |
Matrix of estimated SNVC (= SVC + NVC) on x (N x K) |
bse_vc |
Matrix of standard errors for the SNVCs on x (N x k) |
z_vc |
Matrix of z-values for the SNVCs on x (N x K) |
p_vc |
Matrix of p-values for the SNVCs on x (N x K) |
B_vc_s |
List summarizing estimated SVCs (in SNVC) on x. The four elements are the SVCs (N x K), the standard errors (N x K), z-values (N x K), and p-values (N x K), respectively |
B_vc_n |
List summarizing estimated NVCs (in SNVC) on x. The four elements are the NVCs (N x K), the standard errors (N x K), z-values (N x K), and p-values (N x K), respectively |
c |
Matrix with columns for the estimated coefficients on xconst, their standard errors, z-values, and p-values (K_c x 4). Effective if xconst_nvc = FALSE |
c_vc |
Matrix of estimated NVCs on xconst (N x K_c). Effective if xconst_nvc = TRUE |
cse_vc |
Matrix of standard errors for the NVCs on xconst (N x K_c). Effective if xconst_nvc = TRUE |
cz_vc |
Matrix of z-values for the NVCs on xconst (N x K_c). Effective if xconst_nvc = TRUE |
cp_vc |
Matrix of p-values for the NVCs on xconst (N x K_c). Effective if xconst_nvc = TRUE |
s |
List of variance parameters in the SNVC (SVC + NVC) on x. The first element is a 2 x K matrix summarizing variance parameters for SVC. The (1, k)-th element is the standard deviation of the k-th SVC, while the (2, k)-th element is the Moran's I value that is scaled to take a value between 0 (no spatial dependence) and 1 (strongest spatial dependence). Based on Griffith (2003), the scaled Moran'I value is interpretable as follows: 0.25-0.50:weak; 0.50-0.70:moderate; 0.70-0.90:strong; 0.90-1.00:marked. The second element of s is the vector of standard deviations of the NVCs |
s_c |
Vector of standard deviations of the NVCs on xconst |
vc |
List indicating whether SVC/NVC are removed or not during the BIC/AIC minimization. 1 indicates not removed (replaced with constant) whreas 0 indicates removed |
e |
Vector whose elements are residual standard error (resid_SE), adjusted conditional R2 (adjR2(cond)), restricted log-likelihood (rlogLik), Akaike information criterion (AIC), and Bayesian information criterion (BIC). When method = "ml", restricted log-likelihood (rlogLik) is replaced with log-likelihood (logLik) |
pred |
Vector of predicted values (N x 1) |
resid |
Vector of residuals (N x 1) |
other |
List of other outputs, which are internally used |
Daisuke Murakami
Muller, S., Scealy, J.L., and Welsh, A.H. (2013) Model selection in linear mixed models. Statistical Science, 28 (2), 136-167.
Murakami, D., Yoshida, T., Seya, H., Griffith, D.A., and Yamagata, Y. (2017) A Moran coefficient-based mixed effects approach to investigate spatially varying relationships. Spatial Statistics, 19, 68-89.
Murakami, D., and Griffith, D.A. (2019). Spatially varying coefficient modeling for large datasets: Eliminating N from spatial regressions. Spatial Statistics, 30, 39-64.
Murakami, D. and Griffith, D.A. (2019) A memory-free spatial additive mixed modeling for big spatial data. Japan Journal of Statistics and Data Science. DOI:10.1007/s42081-019-00063-x.
Murakami, D., and Griffith, D.A. (2020) Balancing spatial and non-spatial variations in varying coefficient modeling: a remedy for spurious correlation. ArXiv.
require(spdep) data(boston) y <- boston.c[, "CMEDV"] x <- boston.c[,c("CRIM", "AGE")] xconst <- boston.c[,c("ZN","DIS","RAD","NOX", "TAX","RM", "PTRATIO", "B")] xgroup <- boston.c[,"TOWN"] coords <- boston.c[,c("LON", "LAT")] ############## SVC modeling1 ################# ######## (SVC on x; Constant coefficients on xconst) #res <- besf_vc(y=y,x=x,xconst=xconst,coords=coords, x_sel = FALSE ) #res #plot_s(res,0) # Spatially varying intercept #plot_s(res,1) # 1st SVC #plot_s(res,2) # 2nd SVC # ######## For large samples (n > 5,000), the following additional learning ######## mitigates an degeneracy/over-smoothing problem in SVCs #res1 <- addlearn_local(res) #res1 #plot_s(res1,0) # Spatially varying intercept #plot_s(res1,1) # 1st SVC #plot_s(res1,2) # 2nd SVC ############## SVC modeling2 ################# ######## (SVC or constant coefficients on x; Constant coefficients on xconst) #res2 <- besf_vc(y=y,x=x,xconst=xconst,coords=coords ) ############## SVC modeling3 ################# ######## - Group-level SVC or constant coefficients on x ######## - Constant coefficients on xconst #res3 <- besf_vc(y=y,x=x,xconst=xconst,coords=coords, s_id=xgroup) ############## SNVC modeling1 ################# ######## - SNVC, SVC, NVC, or constant coefficients on x ######## - Constant coefficients on xconst #res4 <- besf_vc(y=y,x=x,xconst=xconst,coords=coords, x_nvc =TRUE) ############## SNVC modeling2 ################# ######## - SNVC, SVC, NVC, or constant coefficients on x ######## - NVC or Constant coefficients on xconst #res5 <- besf_vc(y=y,x=x,xconst=xconst,coords=coords, x_nvc =TRUE, xconst_nvc=TRUE) #plot_s(res5,0) # Spatially varying intercept #plot_s(res5,1) # 1st SNVC (SVC + NVC) #plot_s(res5,1,btype="svc")# SVC in the 1st SNVC #plot_n(res5,1,xtype="x") # NVC in the 1st NVC on x #plot_n(res5,6,xtype="xconst")# NVC in the 6t NVC on xcnostrequire(spdep) data(boston) y <- boston.c[, "CMEDV"] x <- boston.c[,c("CRIM", "AGE")] xconst <- boston.c[,c("ZN","DIS","RAD","NOX", "TAX","RM", "PTRATIO", "B")] xgroup <- boston.c[,"TOWN"] coords <- boston.c[,c("LON", "LAT")] ############## SVC modeling1 ################# ######## (SVC on x; Constant coefficients on xconst) #res <- besf_vc(y=y,x=x,xconst=xconst,coords=coords, x_sel = FALSE ) #res #plot_s(res,0) # Spatially varying intercept #plot_s(res,1) # 1st SVC #plot_s(res,2) # 2nd SVC # ######## For large samples (n > 5,000), the following additional learning ######## mitigates an degeneracy/over-smoothing problem in SVCs #res1 <- addlearn_local(res) #res1 #plot_s(res1,0) # Spatially varying intercept #plot_s(res1,1) # 1st SVC #plot_s(res1,2) # 2nd SVC ############## SVC modeling2 ################# ######## (SVC or constant coefficients on x; Constant coefficients on xconst) #res2 <- besf_vc(y=y,x=x,xconst=xconst,coords=coords ) ############## SVC modeling3 ################# ######## - Group-level SVC or constant coefficients on x ######## - Constant coefficients on xconst #res3 <- besf_vc(y=y,x=x,xconst=xconst,coords=coords, s_id=xgroup) ############## SNVC modeling1 ################# ######## - SNVC, SVC, NVC, or constant coefficients on x ######## - Constant coefficients on xconst #res4 <- besf_vc(y=y,x=x,xconst=xconst,coords=coords, x_nvc =TRUE) ############## SNVC modeling2 ################# ######## - SNVC, SVC, NVC, or constant coefficients on x ######## - NVC or Constant coefficients on xconst #res5 <- besf_vc(y=y,x=x,xconst=xconst,coords=coords, x_nvc =TRUE, xconst_nvc=TRUE) #plot_s(res5,0) # Spatially varying intercept #plot_s(res5,1) # 1st SNVC (SVC + NVC) #plot_s(res5,1,btype="svc")# SVC in the 1st SNVC #plot_n(res5,1,xtype="x") # NVC in the 1st NVC on x #plot_n(res5,6,xtype="xconst")# NVC in the 6t NVC on xcnost
This function evaluates the marginal effects from x (dy/dx) based on the estimation result of resf. This funtion is for non-Gaussian models transforming y using nongauss_y.
coef_marginal( mod )coef_marginal( mod )
mod |
Output from |
b |
Marginal effects from x (dy/dx) |
This function evaluates the marginal effects from x (dy/dx) based on the estimation result of resf_vc. This funtion is for non-Gaussian models transforming y using nongauss_y.
coef_marginal_vc( mod )coef_marginal_vc( mod )
mod |
Output from |
b_vc |
Matrix of the marginal effects of x (dy/dx) (N x K) |
B_vc_n |
Matrix of the sub-marginal effects of x explained by the spatially varying coefficients (N x K) |
B_vc_s |
Matrix of the sub-marginal effects explained by the non-spatially varying coefficients (N x K) |
c |
Matrix of the marginal effects of xconst (N x K_const) |
other |
List of other outputs, which are internally used |
This function estimates the linear eigenvector spatial filtering (ESF) model. The eigenvectors are selected by a forward stepwise method.
esf( y, x = NULL, vif = NULL, meig, fn = "r2" )esf( y, x = NULL, vif = NULL, meig, fn = "r2" )
y |
Vector of explained variables (N x 1) |
x |
Matrix of explanatory variables (N x K). Default is NULL |
vif |
Maximum acceptable value of the variance inflation factor (VIF) (scalar). For example, if vif = 10, eigenvectors are selected so that the maximum VIF value among explanatory variables and eigenvectors is equal to or less than 10. Default is NULL |
meig |
Moran eigenvectors and eigenvalues. Output from |
fn |
Objective function for the stepwise eigenvector selection. The adjusted R2 ("r2"), AIC ("aic"), or BIC ("bic") are available. Alternatively, all the eigenvectors in meig are used without the stepwise selection if fn = "all". This is acceptable for large samples (see Murakami and Griffith, 2019). Default is "r2" |
b |
Matrix with columns for the estimated coefficients on x, their standard errors, t-values, and p-values (K x 4) |
s |
Vector of statistics for the estimated spatial component (2 x 1). The first element is the standard deviation and the second element is the Moran's I value of the estimated spatially dependent component. The Moran's I value is scaled to take a value between 0 (no spatial dependence) and 1 (the maximum possible spatial dependence). Based on Griffith (2003), the scaled Moran'I value is interpretable as follows: 0.25-0.50:weak; 0.50-0.70:moderate; 0.70-0.90:strong; 0.90-1.00:marked |
r |
Matrix with columns for the estimated coefficients on Moran's eigenvectors, their standard errors, t-values, and p-values (L x 4) |
vif |
Vector of variance inflation factors of the explanatory variables (N x 1) |
e |
Vector whose elements are residual standard error (resid_SE), adjusted R2 (adjR2), log-likelihood (logLik), AIC, and BIC |
sf |
Vector of estimated spatial dependent component (E |
pred |
Vector of predicted values (N x 1) |
resid |
Vector of residuals (N x 1) |
other |
List of other outputs, which are internally used |
Daisuke Murakami
Griffith, D. A. (2003). Spatial autocorrelation and spatial filtering: gaining understanding through theory and scientific visualization. Springer Science & Business Media.
Tiefelsdorf, M., and Griffith, D. A. (2007). Semiparametric filtering of spatial autocorrelation: the eigenvector approach. Environment and Planning A, 39 (5), 1193-1221.
Murakami, D. and Griffith, D.A. (2019) Eigenvector spatial filtering for large data sets: fixed and random effects approaches. Geographical Analysis, 51 (1), 23-49.
require(spdep) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE")] coords <- boston.c[,c("LON", "LAT")] #########Distance-based ESF meig <- meigen(coords=coords) esfD <- esf(y=y,x=x,meig=meig, vif=5) esfD #########Fast approximation meig_f<- meigen_f(coords=coords) esfD <- esf(y=y,x=x,meig=meig_f, vif=10, fn="all") esfD ############################Not run #########Topoligy-based ESF (it is commonly used in regional science) # #cknn <- knearneigh(coordinates(coords), k=4) #4-nearest neighbors #cmat <- nb2mat(knn2nb(cknn), style="B") #meig <- meigen(cmat=cmat, threshold=0.25) #esfT <- esf(y=y,x=x,meig=meig) #esfTrequire(spdep) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE")] coords <- boston.c[,c("LON", "LAT")] #########Distance-based ESF meig <- meigen(coords=coords) esfD <- esf(y=y,x=x,meig=meig, vif=5) esfD #########Fast approximation meig_f<- meigen_f(coords=coords) esfD <- esf(y=y,x=x,meig=meig_f, vif=10, fn="all") esfD ############################Not run #########Topoligy-based ESF (it is commonly used in regional science) # #cknn <- knearneigh(coordinates(coords), k=4) #4-nearest neighbors #cmat <- nb2mat(knn2nb(cknn), style="B") #meig <- meigen(cmat=cmat, threshold=0.25) #esfT <- esf(y=y,x=x,meig=meig) #esfT
This function estimates the low rank spatial error model.
lsem( y, x, weig, method = "reml" )lsem( y, x, weig, method = "reml" )
y |
Vector of explained variables (N x 1) |
x |
Matrix of explanatory variables (N x K) |
weig |
eigenvectors and eigenvalues of a spatial weight matrix. Output from |
method |
Estimation method. Restricted maximum likelihood method ("reml") and maximum likelihood method ("ml") are available. Default is "reml" |
b |
Matrix with columns for the estimated coefficients on x, their standard errors, t-values, and p-values (K x 4) |
s |
Vector of estimated variance parameters (2 x 1). The first and the second elements denote the estimated rho parameter (sp_lambda) quantfying the scale of spatial dependent process, and the standard error of the process (sp_SE), respectively. |
e |
Vector whose elements are residual standard error (resid_SE), adjusted conditional R2 (adjR2(cond)), restricted log-likelihood (rlogLik), Akaike information criterion (AIC), and Bayesian information criterion (BIC). When method = "ml", restricted log-likelihood (rlogLik) is replaced with log-likelihood (logLik) |
r |
Vector of estimated random coefficients on the spatial eigenvectors (L x 1) |
pred |
Vector of predicted values (N x 1) |
resid |
Vector of residuals (N x 1) |
other |
List of other outputs, which are internally used |
Daisuke Murakami
Murakami, D., Seya, H. and Griffith, D.A. (2018) Low rank spatial econometric models. Arxiv.
require(spdep) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] coords<- boston.c[,c("LON", "LAT")] weig <- weigen( coords ) res <- lsem(y=y,x=x,weig=weig) resrequire(spdep) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] coords<- boston.c[,c("LON", "LAT")] weig <- weigen( coords ) res <- lsem(y=y,x=x,weig=weig) res
This function estimates the low rank spatial lag model.
lslm( y, x, weig, method = "reml", boot = FALSE, iter = 200 )lslm( y, x, weig, method = "reml", boot = FALSE, iter = 200 )
y |
Vector of explained variables (N x 1) |
x |
Matrix of explanatory variables (N x K) |
weig |
eigenvectors and eigenvalues of a spatial weight matrix. Output from |
method |
Estimation method. Restricted maximum likelihood method ("reml") and maximum likelihood method ("ml") are available. Default is "reml" |
boot |
If it is TRUE, confidence intervals for the spatial dependence parameters (s), the mean direct effects (de), and the mean indirect effects (ie), are estimated through a parametric bootstrapping. Default is FALSE |
iter |
The number of bootstrap replicates. Default is 200 |
b |
Matrix with columns for the estimated coefficients on x, their standard errors, t-values, and p-values (K x 4) |
s |
Vector of estimated shrinkage parameters (2 x 1). The first and the second elements denote the estimated rho parameter (sp_rho) quantfying the scale of spatial dependence, and the standard error of the spatial dependent component (sp_SE), respectively. If boot = TRUE, their 95 percent confidence intervals and the resulting p-values are also provided |
e |
Vector whose elements are residual standard error (resid_SE), adjusted conditional R2 (adjR2(cond)), restricted log-likelihood (rlogLik), Akaike information criterion (AIC), and Bayesian information criterion (BIC). When method = "ml", restricted log-likelihood (rlogLik) is replaced with log-likelihood (logLik) |
de |
Matrix with columns for the estimated mean direct effects on x. If boot = TRUE, their 95 percent confidence intervals and the resulting p-values are also provided |
ie |
Matrix with columns for the estimated mean indirect effects on x. If boot = TRUE, their 95 percent confidence intervals and the resulting p-values are also provided |
r |
Vector of estimated random coefficients on the spatial eigenvectors (L x 1) |
pred |
Vector of predicted values (N x 1) |
resid |
Vector of residuals (N x 1) |
other |
List of other outputs, which are internally used |
Daisuke Murakami
Murakami, D., Seya, H. and Griffith, D.A. (2018) Low rank spatial econometric models. Arxiv.
require(spdep) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] coords <- boston.c[,c("LON", "LAT")] weig <- weigen(coords) res <- lslm(y=y,x=x,weig=weig) ## res <- lslm(y=y,x=x,weig=weig, boot=TRUE) resrequire(spdep) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] coords <- boston.c[,c("LON", "LAT")] weig <- weigen(coords) res <- lslm(y=y,x=x,weig=weig) ## res <- lslm(y=y,x=x,weig=weig, boot=TRUE) res
This function extracts spatial and temporal eigenvectors (i.e., basis functions describing spatial and temporal patterns).
meigen( coords = NULL,model = "exp", enum = NULL, s_id = NULL, threshold = 0, cmat = NULL, coords_z=NULL, enum_z=NULL, interact=TRUE, interact_max_dim = 600 )meigen( coords = NULL,model = "exp", enum = NULL, s_id = NULL, threshold = 0, cmat = NULL, coords_z=NULL, enum_z=NULL, interact=TRUE, interact_max_dim = 600 )
coords |
Matrix of spatial coordinates (N x 2). If cmat is specified, it is ignored |
model |
Type of kernel to model spatial dependence. The currently available options are "exp" for the exponential kernel, "gau" for the Gaussian kernel, and "sph" for the spherical kernel. Default is "exp" |
enum |
Optional. The maximum mumber of spatial eigenvectors to be extracted (scalar) |
s_id |
Optional. Location/zone ID for modeling inter-group spatial effects. If specified, Moran eigenvectors are extracted by groups. It is useful e.g. for multilevel modeling (s_id is the groups) and panel data modeling (s_id is given by individual location id). Default is NULL |
threshold |
Optional. Threshold for the eigenvalues. Suppose that lambda_1 is the maximum eigenvalue, this function extracts eigenvectors whose corresponding eigenvalue is equal or greater than (threshold x lambda_1). threshold must be a value between 0 and 1. Default is zero (see Details) |
cmat |
Optional. A user-specified spatial connectivity matrix (N x N). It must be provided when the user wants to use a spatial connectivity matrix other than the default matrices |
coords_z |
Optional. One- or two-column matrix whose t-th column represents t-th temporal coordinate (N x 1 or N x 2). |
enum_z |
Optional. The maximum mumber of temporal eigenvectors to be extracted (scalar) |
interact |
Optional. If TRUE, space-time eigenvectors (space x time) are considered in addition to spatial eigenvectors and temporal eigenvectors |
interact_max_dim |
Optional. The maximum mumber of the space-time eigenvectors to be extracted (scalar) |
This function extracts spatial eigenvectors from MCM, where M = I - 11'/N is a centering operator. By default, C is a N x N connectivity matrix whose (i, j)-th element equals exp(-d(i,j)/h), where d(i,j) is the spatial Euclidean distance between the sample sites i and j. h is the maximum length of the minimum spanning tree connecting sample sites (see Dray et al., 2006). If cmat is provided, this function performs the same calculation after C is replaced with cmat.
The temporal eigenvectors are extracted in the same way where the spatial distance d(i,j) is replaced with temporal difference. If two temporal coordinates are given, their eigenvectors are evaluated respectively.
If threshold = 0.00 (default), all the eigenvectors corresponding to positive eigenvalues explaining positive spatial/temporal dependence are extracted. threshold = 0.00 or 0.25 are standard assumptions (see Griffith, 2003; Murakami and Griffith, 2015).
sf |
Matrix of the spatial eigenvectors (N x L) |
ev |
Vector of the spatial eigenvalues (L x 1), scaled to have the maximum value of 1 |
sf_z |
List. t-th element is the matrix of the t-th temporal eigenvectors (N x L_t) |
ev_z |
List. t-th element is the vector of the t-th temporal eigenvalues (L_t x 1), scaled to have the maximum value of 1 |
other |
List of other outcomes, which are internally used |
Daisuke Murakami
Dray, S., Legendre, P., and Peres-Neto, P.R. (2006) Spatial modelling: a comprehensive framework for principal coordinate analysis of neighbour matrices (PCNM). Ecological Modelling, 196 (3), 483-493.
Griffith, D.A. (2003) Spatial autocorrelation and spatial filtering: gaining understanding through theory and scientific visualization. Springer Science & Business Media.
Murakami, D. and Griffith, D.A. (2015) Random effects specifications in eigenvector spatial filtering: a simulation study. Journal of Geographical Systems, 17 (4), 311-331.
Murakami, D., Shirota, S., Kajita, S., and Kajita, S. (2024) Fast spatio-temporally varying coefficient modeling with reluctant interaction selection. ArXiv.
meigen_f for fast eigen-decomposition
This function approximates spatial and temporal eigenvectors (i.e., basis functions describing spatial and temporal patterns) computationally efficiently.
meigen_f( coords, model = "exp", enum = 200, s_id = NULL, threshold = 0, coords_z = NULL, enum_z = 200, interact = TRUE, interact_max_dim = 600 )meigen_f( coords, model = "exp", enum = 200, s_id = NULL, threshold = 0, coords_z = NULL, enum_z = 200, interact = TRUE, interact_max_dim = 600 )
coords |
Matrix of spatial coordinates (N x 2) |
model |
Type of kernel to model spatial dependence. The currently available options are "exp" for the exponential kernel, "gau" for the Gaussian kernel, and "sph" for the spherical kernel. Default is "exp" |
enum |
Number of eigenvectors and eigenvalues to be extracted (scalar). Default is 200 |
s_id |
Optional. Location/zone ID for modeling inter-group spatial effects. If specified, Moran eigenvectors are extracted by groups. It is useful e.g. for multilevel modeling (s_id is the groups) and panel data modeling (s_id is given by individual location id). Default is NULL |
threshold |
Optional. Threshold for the eigenvalues. Suppose that lambda_1 is the maximum eigenvalue, this function extracts eigenvectors whose corresponding eigenvalue is equal or greater than (threshold x lambda_1). threshold must be a value between 0 and 1. Default is zero |
coords_z |
Optional. One- or two-column matrix of temporal coordinates (N x 1 or N x 2). |
enum_z |
Optional. The maximum mumber of temporal eigenvectors to be extracted (scalar) |
interact |
Optional. If TRUE, space-time eigenvectors (space x time) are considered in addition to spatial eigenvectors and temporal eigenvectors |
interact_max_dim |
Optional. The maximum mumber of the space-time eigenvectors to be extracted (scalar) |
This function extracts approximated spatial eigenvectors from MCM. M = I - 11'/N is a centering operator, and C is a spatial connectivity matrix whose (i, j)-th element is given by exp( -d(i,j)/h), where d(i,j) is the Euclidean distance between the sample sites i and j, and h is a range parameter given by the maximum length of the minimum spanning tree connecting sample sites (see Dray et al., 2006). Following a simulation result in Murakami and Griffith (2019), this function approximates the 200 eigenvectors corresponding to the 200 largest eigenvalues by default (i.e., enum = 200). If enum is given by a smaller value like 100, the computation time will be shorter, but with greater approximation error.
The temporal eigenvectors are extracted in the same way where the spatial distance d(i,j) is replaced with temporal difference. If two temporal coordinates are given, their eigenvectors are evaluated respectively.
sf |
Matrix of the spatial eigenvectors (N x L) |
ev |
Vector of the spatial eigenvalues (L x 1), scaled to have the maximum value of 1 |
sf_z |
List. t-th element is the matrix of the t-th temporal eigenvectors (N x L_t) |
ev_z |
List. t-th element is the vector of the t-th temporal eigenvalues (L_t x 1), scaled to have the maximum value of 1 |
other |
List of other outcomes, which are internally used |
Daisuke Murakami
Dray, S., Legendre, P., and Peres-Neto, P.R. (2006) Spatial modelling: a comprehensive framework for principal coordinate analysis of neighbour matrices (PCNM). Ecological Modelling, 196 (3), 483-493.
Murakami, D. and Griffith, D.A. (2019) Eigenvector spatial filtering for large data sets: fixed and random effects approaches. Geographical Analysis, 51 (1), 23-49.
Murakami, D., Shirota, S., Kajita, S., and Kajita, S. (2024) Fast spatio-temporally varying coefficient modeling with reluctant interaction selection. ArXiv.
This function estimates Moran eigenvectors at unobserved sites using the Nystrom extension.
meigen0( meig, coords0, coords_z0 = NULL, s_id0 = NULL )meigen0( meig, coords0, coords_z0 = NULL, s_id0 = NULL )
meig |
Moran eigenvectors and eigenvalues. Output from |
coords0 |
Matrix of spatial point coordinates of prediction sites (N_0 x 2) |
coords_z0 |
Optional. One- or two-column matrix whose t-th column represents the t-th temporal coordinate of prediction times (N_0 x 1 or N_0 x 2). |
s_id0 |
Optional. ID specifying groups modeling spatial effects (N_0 x 1). If specified, Moran eigenvectors are extracted by groups. It is useful e.g. for multilevel modeling (s_id is the groups) and panel data modeling (s_id is given by individual location id). Default is NULL |
sf |
Matrix of the first L eigenvectors at unobserved sites (N_0 x L) |
ev |
Vector of the first L eigenvalues (L x 1) |
sf_z |
List. t-th element is the matrix of the t-th temporal eigenvectors (N x L_t) |
ev_z |
List. t-th element is the vector of the t-th temporal eigenvalues (L_t x 1) |
other |
List of other outputs, which are internally used |
Daisuke Murakami
Drineas, P. and Mahoney, M.W. (2005) On the Nystrom method for approximating a gram matrix for improved kernel-based learning. Journal of Machine Learning Research, 6 (2005), 2153-2175.
Parameter setup for modeling non-Gaussian continuous data and count data. The SAL transformation (see details) is used to model a wide variety of non-Gaussian data without explicitly assuming data distribution (see Murakami et al., 2021 for further detail). In addition, Box-Cox transformation is used for non-negative continuous variables while another transformation approximating overdispersed Poisson distribution is used for count variables. The output from this function is used as an input of the resf and resf_vc functions. For further details about its implementation and case study examples, see Murakami (2021).
nongauss_y( y_type = "continuous", y_nonneg = FALSE, tr_num = 0 )nongauss_y( y_type = "continuous", y_nonneg = FALSE, tr_num = 0 )
y_type |
Type of explained variables y. "continuous" for continuous variables and "count" for count variables |
y_nonneg |
Effective if y_type = "continuous". TRUE if y cannot take negative value. If y_nonneg = TRUE and tr_num = 0, the Box-Cox transformation is applied to y. If y_nonneg = TRUE and tr_num > 0, the Box-Cox transformation is applied first to roughly Gaussianize y. Then, the SAL transformation is iterated tr_num times to improve the modeling accuracy. Default is FALSE |
tr_num |
Number of the SAL transformations (SinhArcsinh and Affine, where the use of "L" stems from the "Linear") applied to Gaussianize y. Default is 0 |
If tr_num >0, the SAL transformation is iterated tr_num times to Gaussianize y. The SAL transformation is defined as SAL(y)=a+b*sinh(c*arcsinh(y)-d) where a,b,c,d are parameters. Based on Rios and Tobar (2019), the iteration of the SAL transformation approximates a wide variety of non-Gaussian distributions without explicitly assuming data distribution. The resf and resf_vc functions return tr_par, which is a list whose k-th element includes the a,b,c,d parameters used for the k-th SAL transformation.
In addition, for non-negative y (y_nonneg = TRUE), the Box-Cox transformation is applied prior to the iterative SAL transformation. tr_num and y_nonneg can be selected by comparing the BIC (or AIC) values across models. This compositionally-warped spatial regression approach is detailed in Murakami et al. (2021).
For count data (y_type = "count"), an overdispersed Poisson distribution (Gaussian approximation) is assumed. If tr_num > 0, the distribution is adjusted to fit the data (y) through the iterative SAL transformations. y_nonneg is ignored if y_type = "count".
nongauss |
List of parameters for modeling non-Gaussian data |
Rios, G. and Tobar, F. (2019) Compositionally-warped Gaussian processes. Neural Networks, 118, 235-246.
Murakami, D. (2021) Transformation-based generalized spatial regression using the spmoran package: Case study examples, ArXiv.
Murakami, D., Kajita, M., Kajita, S. and Matsui, T. (2021) Compositionally-warped additive mixed modeling for a wide variety of non-Gaussian data. Spatial Statistics, 43, 100520.
Murakami, D., & Matsui, T. (2021). Improved log-Gaussian approximation for over-dispersed Poisson regression: application to spatial analysis of COVID-19. ArXiv, 2104.13588.
###### Regression for non-negative data (BC trans.) ng1 <-nongauss_y( y_nonneg = TRUE ) ng1 ###### General non-Gaussian regression for continuous data (two SAL trans.) ng2 <-nongauss_y( tr_num = 2 ) ng2 ###### General non-Gaussian regression for non-negative continuous data ng3 <-nongauss_y( y_nonneg = TRUE, tr_num = 5 ) ng3 ###### Over-dispersed Poisson regression for count data ng4 <-nongauss_y( y_type = "count" ) ng4 ###### A general non-Gaussian regression for count data ng5 <-nongauss_y( y_type = "count", tr_num = 5 ) ng5 ############################## Fitting example require(spdep);require(Matrix) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] xgroup<- boston.c[,"TOWN"] coords<- boston.c[,c("LON","LAT")] meig <- meigen(coords=coords) res <- resf(y = y, x = x, meig = meig,nongauss=ng2) res # Estimation results plot(res$pdf,type="l") # Estimated probability density function res$skew_kurt # Skew and kurtosis of the estimated PDF res$pred_quantile[1:2,]# predicted value by quantile coef_marginal(res) # Estimated marginal effects (dy/dx)###### Regression for non-negative data (BC trans.) ng1 <-nongauss_y( y_nonneg = TRUE ) ng1 ###### General non-Gaussian regression for continuous data (two SAL trans.) ng2 <-nongauss_y( tr_num = 2 ) ng2 ###### General non-Gaussian regression for non-negative continuous data ng3 <-nongauss_y( y_nonneg = TRUE, tr_num = 5 ) ng3 ###### Over-dispersed Poisson regression for count data ng4 <-nongauss_y( y_type = "count" ) ng4 ###### A general non-Gaussian regression for count data ng5 <-nongauss_y( y_type = "count", tr_num = 5 ) ng5 ############################## Fitting example require(spdep);require(Matrix) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] xgroup<- boston.c[,"TOWN"] coords<- boston.c[,c("LON","LAT")] meig <- meigen(coords=coords) res <- resf(y = y, x = x, meig = meig,nongauss=ng2) res # Estimation results plot(res$pdf,type="l") # Estimated probability density function res$skew_kurt # Skew and kurtosis of the estimated PDF res$pred_quantile[1:2,]# predicted value by quantile coef_marginal(res) # Estimated marginal effects (dy/dx)
This function plots non-spatially varying coefficients (NVCs; coefficients varying with respect to explanatory variable value) and their 95 percent confidence intervals
plot_n( mod, xnum = 1, xtype = "x", cex.lab = 20, cex.axis = 15, lwd = 1.5, ylim = NULL, nmax = 20000 )plot_n( mod, xnum = 1, xtype = "x", cex.lab = 20, cex.axis = 15, lwd = 1.5, ylim = NULL, nmax = 20000 )
mod |
|
xnum |
The NVC on the xnum-th explanatory variable is plotted. Default is 1 |
xtype |
Effective for |
cex.lab |
The size of the x and y axis labels |
cex.axis |
The size of the tick label numbers |
lwd |
The width of the line drawing the coefficient estimates |
ylim |
The limints of the y-axis |
nmax |
If sample size exceeds nmax, nmax samples are randomly selected and plotted. Default is 20,000 |
This function plots regression coefficients estimated from the spatial filter unconditional quantile regression (SF-UQR) model.
plot_qr( mod, pnum = 1, par = "b", cex.main = 20, cex.lab = 18, cex.axis = 15, lwd = 1.5 )plot_qr( mod, pnum = 1, par = "b", cex.main = 20, cex.lab = 18, cex.axis = 15, lwd = 1.5 )
mod |
Outpot from the |
pnum |
A number specifying the parameter being plotted. If par = "b", the coefficients on the pnum-th explanatory variable are plotted (intercepts are plotted if pnum = 1). If par = "s" and pnum = 1, the estimated standard errors for the reidual spatial process are plotted. If par = "s" and pnum = 2, the Moran's I values of the residual spatial process are plotted. The Moran's I value is scaled to take a value between 0 (no spatial dependence) and 1 (the maximum possible spatial dependence). Based on Griffith (2003), the scaled Moran'I value is interpretable as follows: 0.25-0.50:weak; 0.50-0.70:moderate; 0.70-0.90:strong; 0.90-1.00:marked |
par |
If it is "b", regression coefficeints are plotted. If it is "s", shrinkage (variance) parameters for the residual spatial process are plotted. Default is "b" |
cex.main |
Graphical parameter specifying the size of the main title |
cex.lab |
Graphical parameter specifying the size of the x and y axis labels |
cex.axis |
Graphical parameter specifying the size of the tick label numbers |
lwd |
Graphical parameters specifying the width of the line drawing the coefficient estimates |
See par for the graphical parameters
This function plots spatially varying coefficients (SVC) and spatio-temporally varying coefficients (STVC) with/without coefficient varying with respect to the value of the explanatory variable (NVC). Namely, the full varying coefficient equals STVC + NVC.
plot_s( mod, xnum = 0, btype = "all", xtype = "x", pmax = NULL, ncol = 8, col = NULL, inv =FALSE, brks = "regular", cex = 1, pch = 20, nmax = 20000, coords_z1_lim=NULL, coords_z2_lim = NULL)plot_s( mod, xnum = 0, btype = "all", xtype = "x", pmax = NULL, ncol = 8, col = NULL, inv =FALSE, brks = "regular", cex = 1, pch = 20, nmax = 20000, coords_z1_lim=NULL, coords_z2_lim = NULL)
mod |
Outpot from |
xnum |
For |
btype |
Effective if x_nvc =TRUE in |
xtype |
If "x" (default), coefficients on x is plotted. If "xconst", those on xconst is plotted |
pmax |
The maximum p-value for the varying coefficients to be displayed. For example, if pmax = 0.05, the only coefficients that are statistically significant at the 5 percent level are plotted. If NULL, all the coefficients are plotted. Default is NULL |
ncol |
Number of colors in the color palette. Default is 8 |
col |
Color palette used for the mapping. If NULL, the blue-pink-yellow color scheme is used. Palettes in the RColorBrewer package are available. Default is NULL |
inv |
If TRUE, the color palett is inverted. Default is FALSE |
brks |
If "regular", color is changed at regular intervals. If "quantile", color is changed for each quantile |
cex |
Size of the dots representing sample sites |
pch |
A number indicating the symbol to use |
nmax |
If sample size exceeds nmax, nmax samples are randomly selected and plotted. Default is 20,000 |
coords_z1_lim |
Value range for coords_z[,1] in the |
coords_z2_lim |
Value range for coords_z[,2] (vector). |
resf, besf, resf_vc, besf_vc, addlearn_local
It is a function for spatial/spatio-temporal pprediction using the model estimated from esf, resf, or resf_vc function.
predict0( mod, meig0, x0 = NULL, xconst0 = NULL, xgroup0 = NULL, offset0 = NULL, weight0 = NULL, compute_se=FALSE, compute_quantile = FALSE )predict0( mod, meig0, x0 = NULL, xconst0 = NULL, xgroup0 = NULL, offset0 = NULL, weight0 = NULL, compute_se=FALSE, compute_quantile = FALSE )
mod |
, or resf_vc
meig0 |
Moran eigenvectors at prediction sites. Output from |
x0 |
Matrix of explanatory variables at prediction sites (N_0 x K). Each column of x0 must correspond to those in x in the input model (mod). Default is NULL |
xconst0 |
Effective for |
xgroup0 |
Matrix/vector of group IDs at prediction sites that may be integer or name by group (N_0 x K_g). Default is NULL |
offset0 |
Vector of offset variables at prediction sites (N_0 x 1). Effective if y is count (see |
weight0 |
Vector of weights for prediction sites (N_0 x 1). Required if compute_se = TRUE or compute_quantile = TRUE, and weight in the input model is not NULL |
compute_se |
If TRUE, predictive standard error is evaulated. It is currently supported only for continuous variables. If nongauss is specified in the input model (mod), standard error for the transformed y is evaluated. Default is FALSE |
compute_quantile |
If TRUE, Matrix of the quantiles for the predicted values (N x 15) is evaulated. It is currently supported only for continuous variables. Default is FALSE |
pred |
Matrix with the first column for the predicted values (pred). The second and the third columns are the predicted trend component (xb) and the residual spatial process (sf_residual). If xgroup0 is specified, the fourth column is the predicted group effects (group). If tr_num > 0 or tr_nonneg ==TRUE (i.e., y is transformed) in mod, there is another column of the predicted values in the transformed/normalized scale (pred_trans). In addition, if compute_quantile =TRUE, predictive standard error (pred_se) is evaluated and added as another column |
pred_quantile |
Effective if compute_quantile = TRUE. Matrix of the quantiles for the predicted values (N x 15). It is useful for evaluating uncertainty in the predictive values |
b_vc |
Matrix of estimated spatially (spatio-temporally) varying coefficients (S(T)VCs) on x0 (N_0 x K) |
bse_vc |
Matrix of estimated standard errors for the S(T)VCs (N_0 x K) |
t_vc |
Matrix of estimated t-values for the S(T)VCs (N_0 x K) |
p_vc |
Matrix of estimated p-values for the S(T)VCs (N_0 x K) |
c_vc |
Matrix of estimated non-spatially varying coefficients (NVCs) on x0 (N x K). Effective if nvc =TRUE in |
cse_vc |
Matrix of standard errors for the NVCs on x0 (N x K).Effective if nvc =TRUE in |
ct_vc |
Matrix of t-values for the NVCs on x0 (N x K). Effective if nvc =TRUE in |
cp_vc |
Matrix of p-values for the NVCs on x0 (N x K). Effective if nvc =TRUE in |
require(spdep) data(boston) samp <- sample( dim( boston.c )[ 1 ], 300) d <- boston.c[ samp, ] ## Data at observed sites y <- d[, "CMEDV"] x <- d[,c("ZN", "LSTAT")] xconst <- d[,c("CRIM", "NOX", "AGE", "DIS", "RAD", "TAX", "PTRATIO", "B", "RM")] coords <- d[,c("LON", "LAT")] d0 <- boston.c[-samp, ] ## Data at unobserved sites y0 <- d0[, "CMEDV"] x0 <- d0[,c("ZN", "LSTAT")] xconst0 <- d0[,c("CRIM", "NOX", "AGE", "DIS", "RAD", "TAX", "PTRATIO", "B", "RM")] coords0 <- d0[,c("LON", "LAT")] meig <- meigen( coords = coords ) meig0 <- meigen0( meig = meig, coords0 = coords0 ) ############ Spatial prediction ############ #### model with residual spatial dependence mod <- resf(y=y, x=x, meig=meig) pred0 <- predict0( mod = mod, x0 = x0, meig0 = meig0 ) pred0$pred[1:5,] # Predicted values #### model with spatially varying coefficients (SVCs) mod <- resf_vc(y=y, x=x, xconst=xconst, meig=meig ) pred0 <- predict0( mod = mod, x0 = x0, xconst0=xconst0, meig0 = meig0 ) pred0$pred[1:5,] # Predicted values pred0$b_vc[1:5,] # SVCs pred0$bse_vc[1:5,]# standard errors of the SVCs pred0$t_vc[1:5,] # t-values of the SNVCs pred0$p_vc[1:5,] # p-values of the SNVCs plot(y0,pred0$pred[,1]);abline(0,1)require(spdep) data(boston) samp <- sample( dim( boston.c )[ 1 ], 300) d <- boston.c[ samp, ] ## Data at observed sites y <- d[, "CMEDV"] x <- d[,c("ZN", "LSTAT")] xconst <- d[,c("CRIM", "NOX", "AGE", "DIS", "RAD", "TAX", "PTRATIO", "B", "RM")] coords <- d[,c("LON", "LAT")] d0 <- boston.c[-samp, ] ## Data at unobserved sites y0 <- d0[, "CMEDV"] x0 <- d0[,c("ZN", "LSTAT")] xconst0 <- d0[,c("CRIM", "NOX", "AGE", "DIS", "RAD", "TAX", "PTRATIO", "B", "RM")] coords0 <- d0[,c("LON", "LAT")] meig <- meigen( coords = coords ) meig0 <- meigen0( meig = meig, coords0 = coords0 ) ############ Spatial prediction ############ #### model with residual spatial dependence mod <- resf(y=y, x=x, meig=meig) pred0 <- predict0( mod = mod, x0 = x0, meig0 = meig0 ) pred0$pred[1:5,] # Predicted values #### model with spatially varying coefficients (SVCs) mod <- resf_vc(y=y, x=x, xconst=xconst, meig=meig ) pred0 <- predict0( mod = mod, x0 = x0, xconst0=xconst0, meig0 = meig0 ) pred0$pred[1:5,] # Predicted values pred0$b_vc[1:5,] # SVCs pred0$bse_vc[1:5,]# standard errors of the SVCs pred0$t_vc[1:5,] # t-values of the SNVCs pred0$p_vc[1:5,] # p-values of the SNVCs plot(y0,pred0$pred[,1]);abline(0,1)
This model estimates regression coefficients, coefficients varying depending on x (non-spatially varying coefficients; NVC), and group effects, considering residual spatial/spatio-temporal dependence. The random-effects eigenvector spatial filtering, which is an approximate Gaussian process approach, is used for modeling the residual dependence. If nonugauss is specified, non-Gaussian explained variables are Gaussianized using a compositional warping function (see nongauss_y). This augument allows the resf function to be applied to non-Gaussian explained variables, including count data.
resf( y, x = NULL, xgroup = NULL, weight = NULL, offset = NULL, nvc = FALSE, nvc_sel = TRUE, nvc_num = 5, meig, method = "reml", penalty = "bic", nongauss = NULL )resf( y, x = NULL, xgroup = NULL, weight = NULL, offset = NULL, nvc = FALSE, nvc_sel = TRUE, nvc_num = 5, meig, method = "reml", penalty = "bic", nongauss = NULL )
y |
Vector of explained variables (N x 1) |
x |
Matrix of explanatory variables (N x K). Default is NULL |
xgroup |
Matrix of group IDs. The IDs may be group numbers or group names (N x K_g). Default is NULL |
weight |
Vector of weights for samples (N x 1). If non-NULL, the adjusted R-squared value is evaluated for weighted explained variables. Default is NULL |
offset |
Vector of offset variables (N x 1). Available if y is count (y_type = "count" is specified in the |
nvc |
If TRUE, a non-linear function of x (NVC; a spline function) is used as a varying coefficient. If FALSE, constant coefficients are assumed. Default is FALSE |
nvc_sel |
If TRUE, type of each coefficient (NVC or constant) is selected through a BIC minimization. If FALSE, NVCs are assumed across x. Alternatively, nvc_sel can be given by column number(s) of x. For example, if nvc_sel = 2, the coefficient on the second explanatory variable is NVC and the other coefficients are constants. Default is TRUE |
nvc_num |
Number of natural spline basis functions to be used to model NVC. Default is 5 |
meig |
Moran eigenvectors and eigenvalues. Output from |
method |
Estimation method. Restricted maximum likelihood method ("reml") and maximum likelihood method ("ml") are available. Default is "reml" |
penalty |
Penalty to select type of coefficients (NVC or constant) to stablize the estimates. The current options are "bic" for the Baysian information criterion-type penalty (N x log(K)) and "aic" for the Akaike information criterion (2K). Default is "bic" |
nongauss |
Parameter setup for modeling non-Gaussian continuous data or count data. Output from |
For modeling non-Gaussian data including count data, see nongauss_y.
b |
Matrix with columns for the estimated constant coefficients on x, their standard errors, t-values, and p-values (K x 4) |
b_g |
List of K_g matrices with columns for the estimated group effects, their standard errors, and t-values |
c_vc |
Matrix of estimated NVCs on x (N x K). Effective if nvc = TRUE |
cse_vc |
Matrix of standard errors for the NVCs on x (N x K). Effective if nvc = TRUE |
ct_vc |
Matrix of t-values for the NVCs on x (N x K). Effective if nvc = TRUE |
cp_vc |
Matrix of p-values for the NVCs on x (N x K). Effective if nvc = TRUE |
s |
Vector of estimated variance parameters (2 x 1). The first and the second elements are the standard deviation and the Moran's I value of the estimated spatially (and temporally) dependent process, respectively. The Moran's I value is scaled to take a value between 0 (no spatial dependence) and 1 (the maximum possible spatial dependence). Based on Griffith (2003), the scaled Moran'I value is interpretable as follows: 0.25-0.50:weak; 0.50-0.70:moderate; 0.70-0.90:strong; 0.90-1.00:marked |
s_c |
Vector of standard deviations of the NVCs on xconst |
s_g |
Vector of estimated standard deviations of the group effects |
e |
Error statistics. When y_type="continuous", it includes residual standard error (resid_SE), adjusted conditional R2 (adjR2(cond)), restricted log-likelihood (rlogLik), Akaike information criterion (AIC), and Bayesian information criterion (BIC). rlogLik is replaced with log-likelihood (logLik) if method = "ml". resid_SE is replaced with the residual standard error for the transformed y (resid_SE_trans) if nongauss is specified. When y_type="count", the error statistics contains root mean squared error (RMSE), Gaussian likelihood approximating the model, AIC and BIC based on the likelihood, and the proportion of the null deviance explained by the model (deviance explained (%)). deviance explained, which is also used in the mgcv package, corresponds to the adjusted R2 in case of the linear regression |
vc |
List indicating whether NVC are removed or not during the BIC minimization. 1 indicates not removed whreas 0 indicates removed |
r |
Vector of estimated random coefficients on the Moran's eigenvectors (L x 1) |
sf |
Vector of estimated spatial dependent component (N x 1) |
pred |
Matrix of predicted values for y (pred) and their standard errors (pred_se) (N x 2). If y is transformed by specifying |
pred_quantile |
Matrix of the quantiles for the predicted values (N x 15). It is useful to evaluate uncertainty in the predictive value |
tr_par |
List of the parameter estimates for the tr_num SAL transformations. The k-th element of the list includes the four parameters for the k-th SAL transformation (see |
tr_bpar |
The estimated parameter in the Box-Cox transformation |
tr_y |
Vector of the transformed explaied variables |
resid |
Vector of residuals (N x 1) |
pdf |
Matrix whose first column consists of evenly spaced values within the value range of y and the second column consists of the estimated value of the probability density function for y if y_type in |
skew_kurt |
Skewness and kurtosis of the estimated probability density/mass function of y |
other |
List of other outputs, which are internally used |
Daisuke Murakami
Murakami, D. and Griffith, D.A. (2015) Random effects specifications in eigenvector spatial filtering: a simulation study. Journal of Geographical Systems, 17 (4), 311-331.
Murakami, D., and Griffith, D.A. (2020) Balancing spatial and non-spatial variations in varying coefficient modeling: a remedy for spurious correlation. Geographical Analysis, DOI: 10.1111/gean.12310.
Murakami, D., Kajita, M., Kajita, S. and Matsui, T. (2021) Compositionally-warped additive mixed modeling for a wide variety of non-Gaussian data. Spatial Statistics, 43, 100520.
Murakami, D., Shirota, S., Kajita, S., and Kajita, S. (2024) Fast spatio-temporally varying coefficient modeling with reluctant interaction selection. ArXiv.
meigen, meigen_f, coef_marginal, besf
##################################################### ############ Spatial regression modeling ############ ##################################################### require(spdep);require(Matrix) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] xgroup<- boston.c[,"TOWN"] coords<- boston.c[,c("LON","LAT")] meig <- meigen(coords=coords) # meig<- meigen_f(coords=coords) ## for large samples ##################################################### ####### Gaussian regression ######################### res <- resf(y = y, x = x, meig = meig) res plot_s(res) ## spatially dependent component (intercept) #### Group-wise random intercepts #res2 <- resf(y = y, x = x, meig = meig, xgroup = xgroup) #### Group-level spatial dependence (s_id) + random intercepts (xgroup) #meig_g<- meigen(coords=coords, s_id = xgroup) #res3 <- resf(y = y, x = x, meig = meig_g, xgroup = xgroup) #### Coefficients varying depending on x #res4 <- resf(y = y, x = x, meig = meig, nvc = TRUE) #res4 #plot_s(res4) # spatially dependent component (intercept) #plot_s(res4,5) # spatial plot of the 5-th NVC #plot_s(res4,6) # spatial plot of the 6-th NVC #plot_s(res4,13)# spatial plot of the 13-th NVC #plot_n(res4,5) # 1D plot of the 5-th NVC #plot_n(res4,6) # 1D plot of the 6-th NVC #plot_n(res4,13)# 1D plot of the 13-th NVC ##################################################### ###### Non-Gaussian regression ###################### #### Model for non-Gaussian continuous data # - Probability distribution is estimated from data #ng5 <- nongauss_y( tr_num = 2 )# 2 SAL transformations to Gaussianize y #res5 <- resf(y = y, x = x, meig = meig, nongauss = ng5) #res5 ## tr_num may be selected by comparing BIC #plot(res5$pdf,type="l") # Estimated probability density function #res5$skew_kurt # Skew and kurtosis of the estimated PDF #res5$pred_quantile[1:2,]# predicted value by quantile #coef_marginal(res5) # Estimated marginal effects (dy/dx) #### Model for non-Gaussian and non-negative continuous data # - Probability distribution is estimated from data #ng6 <- nongauss_y( tr_num = 2, y_nonneg = TRUE ) #res6 <- resf(y = y, x = x, meig = meig, nongauss = ng6 ) #coef_marginal(res6) #### Overdispersed Poisson model for count data # - y: count data #ng7 <- nongauss_y( y_type = "count" ) #res7 <- resf(y = y, x = x, meig = meig, nongauss = ng7 ) #### General model for count data # - y: count data # - Probability distribution is estimated from data #ng8 <- nongauss_y( y_type = "count", tr_num = 2 ) #res8 <- resf(y = y, x = x, meig = meig, nongauss = ng8 ) ##################################################### ################ STVC modeling ###################### ##################################################### # See \url{https://github.com/dmuraka/spmoran} #require(spData) #data(house) #dat0 <- st_as_sf(house) #dat <- data.frame(st_coordinates(dat0), dat0) #y <- log(dat[,"price"]) #x <- dat[,c("lotsize","TLA", "rooms","beds")] #byear <- house$yrbuilt #syear <- as.numeric(as.character(house$syear))#factor -> numeric #coords_z<- cbind(byear,syear) #meig <- meigen_f(coords=coords, coords_z=cbind(byear,syear),interact=TRUE) #res9 <- resf(y=y,x=x,meig=meig ) #res9##################################################### ############ Spatial regression modeling ############ ##################################################### require(spdep);require(Matrix) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] xgroup<- boston.c[,"TOWN"] coords<- boston.c[,c("LON","LAT")] meig <- meigen(coords=coords) # meig<- meigen_f(coords=coords) ## for large samples ##################################################### ####### Gaussian regression ######################### res <- resf(y = y, x = x, meig = meig) res plot_s(res) ## spatially dependent component (intercept) #### Group-wise random intercepts #res2 <- resf(y = y, x = x, meig = meig, xgroup = xgroup) #### Group-level spatial dependence (s_id) + random intercepts (xgroup) #meig_g<- meigen(coords=coords, s_id = xgroup) #res3 <- resf(y = y, x = x, meig = meig_g, xgroup = xgroup) #### Coefficients varying depending on x #res4 <- resf(y = y, x = x, meig = meig, nvc = TRUE) #res4 #plot_s(res4) # spatially dependent component (intercept) #plot_s(res4,5) # spatial plot of the 5-th NVC #plot_s(res4,6) # spatial plot of the 6-th NVC #plot_s(res4,13)# spatial plot of the 13-th NVC #plot_n(res4,5) # 1D plot of the 5-th NVC #plot_n(res4,6) # 1D plot of the 6-th NVC #plot_n(res4,13)# 1D plot of the 13-th NVC ##################################################### ###### Non-Gaussian regression ###################### #### Model for non-Gaussian continuous data # - Probability distribution is estimated from data #ng5 <- nongauss_y( tr_num = 2 )# 2 SAL transformations to Gaussianize y #res5 <- resf(y = y, x = x, meig = meig, nongauss = ng5) #res5 ## tr_num may be selected by comparing BIC #plot(res5$pdf,type="l") # Estimated probability density function #res5$skew_kurt # Skew and kurtosis of the estimated PDF #res5$pred_quantile[1:2,]# predicted value by quantile #coef_marginal(res5) # Estimated marginal effects (dy/dx) #### Model for non-Gaussian and non-negative continuous data # - Probability distribution is estimated from data #ng6 <- nongauss_y( tr_num = 2, y_nonneg = TRUE ) #res6 <- resf(y = y, x = x, meig = meig, nongauss = ng6 ) #coef_marginal(res6) #### Overdispersed Poisson model for count data # - y: count data #ng7 <- nongauss_y( y_type = "count" ) #res7 <- resf(y = y, x = x, meig = meig, nongauss = ng7 ) #### General model for count data # - y: count data # - Probability distribution is estimated from data #ng8 <- nongauss_y( y_type = "count", tr_num = 2 ) #res8 <- resf(y = y, x = x, meig = meig, nongauss = ng8 ) ##################################################### ################ STVC modeling ###################### ##################################################### # See \url{https://github.com/dmuraka/spmoran} #require(spData) #data(house) #dat0 <- st_as_sf(house) #dat <- data.frame(st_coordinates(dat0), dat0) #y <- log(dat[,"price"]) #x <- dat[,c("lotsize","TLA", "rooms","beds")] #byear <- house$yrbuilt #syear <- as.numeric(as.character(house$syear))#factor -> numeric #coords_z<- cbind(byear,syear) #meig <- meigen_f(coords=coords, coords_z=cbind(byear,syear),interact=TRUE) #res9 <- resf(y=y,x=x,meig=meig ) #res9
This function estimates the spatial filter unconditional quantile regression (SF-UQR) model.
resf_qr( y, x = NULL, meig, tau = NULL, boot = TRUE, iter = 200, parallel=FALSE, ncores=NULL )resf_qr( y, x = NULL, meig, tau = NULL, boot = TRUE, iter = 200, parallel=FALSE, ncores=NULL )
y |
Vector of explained variables (N x 1) |
x |
Matrix of explanatory variables (N x K). Default is NULL |
meig |
Moran eigenvectors and eigenvalues. Output from |
tau |
The quantile(s) to be modeled. It must be a number (or a vector of numbers) strictly between 0 and 1. By default, tau = c(0.1, 0.2, ..., 0.9) |
boot |
If it is TRUE, confidence intervals of regression coefficients are estimated by a semiparametric bootstrapping. Default is TRUE |
iter |
The number of bootstrap replications. Default is 200 |
parallel |
If TRUE, the bootstrapping for estimating confidence intervals is parallelized. Default is FALSE |
ncores |
Number of cores used for the parallel computation. If ncores=NULL, which is the default, the number of available cores - 2 is detected and used |
b |
Matrix of estimated regression coefficients (K x Q), where Q is the number of quantiles (i.e., the length of tau) |
r |
Matrix of estimated random coefficients on Moran eigenvectors (L x Q) |
s |
Vector of estimated variance parameters (2 x 1). The first and the second elements denote the standard deviation and the Moran's I value of the estimated spatially dependent component, respectively. The Moran's I value is scaled to take a value between 0 (no spatial dependence) and 1 (the maximum possible spatial dependence). Based on Griffith (2003), the scaled Moran'I value is interpretable as follows: 0.25-0.50:weak; 0.50-0.70:moderate; 0.70-0.90:strong; 0.90-1.00:marked |
e |
Vector whose elements are residual standard error (resid_SE) and adjusted quasi conditional R2 (quasi_adjR2(cond)) |
B |
Q matrices (K x 4) summarizing bootstrapped estimates for the regression coefficients. Columns of these matrices consist of the estimated coefficients, the lower and upper bounds for the 95 percent confidencial intervals, and p-values. It is returned if boot = TRUE |
S |
Q matrices (2 x 3) summarizing bootstrapped estimates for the variance parameters. Columns of these matrices consist of the estimated parameters, the lower and upper bounds for the 95 percent confidencial intervals. It is returned if boot = TRUE |
B0 |
List of Q matrices (K x iter) summarizing bootstrapped coefficients. The q-th matrix consists of the coefficients on the q-th quantile. Effective if boot = TRUE |
S0 |
List of Q matrices (2 x iter) summarizing bootstrapped variance parameters. The q-th matrix consists of the parameters on the q-th quantile. Effective if boot = TRUE |
Daisuke Murakami
Murakami, D. and Seya, H. (2017) Spatially filtered unconditional quantile regression. ArXiv.
require(spdep) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] coords <- boston.c[,c("LON", "LAT")] meig <- meigen(coords=coords) res <- resf_qr(y=y,x=x,meig=meig, boot=FALSE) res plot_qr(res,1) # Intercept plot_qr(res,2) # Coefficient on CRIM plot_qr(res,1,"s") # spcomp_SE plot_qr(res,2,"s") # spcomp_Moran.I/max(Moran.I) ###Not run #res <- resf_qr(y=y,x=x,meig=meig, boot=TRUE) #res #plot_qr(res,1) # Intercept + 95 percent confidence interval (CI) #plot_qr(res,2) # Coefficient on CRIM + 95 percent CI #plot_qr(res,1,"s") # spcomp_SE + 95 percent CI #plot_qr(res,2,"s") # spcomp_Moran.I/max(Moran.I) + 95 percent CIrequire(spdep) data(boston) y <- boston.c[, "CMEDV" ] x <- boston.c[,c("CRIM","ZN","INDUS", "CHAS", "NOX","RM", "AGE", "DIS" ,"RAD", "TAX", "PTRATIO", "B", "LSTAT")] coords <- boston.c[,c("LON", "LAT")] meig <- meigen(coords=coords) res <- resf_qr(y=y,x=x,meig=meig, boot=FALSE) res plot_qr(res,1) # Intercept plot_qr(res,2) # Coefficient on CRIM plot_qr(res,1,"s") # spcomp_SE plot_qr(res,2,"s") # spcomp_Moran.I/max(Moran.I) ###Not run #res <- resf_qr(y=y,x=x,meig=meig, boot=TRUE) #res #plot_qr(res,1) # Intercept + 95 percent confidence interval (CI) #plot_qr(res,2) # Coefficient on CRIM + 95 percent CI #plot_qr(res,1,"s") # spcomp_SE + 95 percent CI #plot_qr(res,2,"s") # spcomp_Moran.I/max(Moran.I) + 95 percent CI
This function estimates spatially varying coefficients (SVC) or spatio-temporally varying coefficients (STVC), group effects, considering residual spatial/spatio-temporal dependence. A non-linear function of x (NVC) can be added on each SVC/STVC mainly to stablize the estimation (see Murakami and Griffith, 2020). Approximate Gaussian processes based on Moran eigenvectors are used for modeling the spatio-temporal processes. Type of coefficients (constant or varying) is selected through a BIC minimization. If nonugauss is specified, non-Gaussian explained variables are Gaussianized using a compositional warping function (see nongauss_y). This augument allows the resf function to be applied to non-Gaussian explained variables, including count data.
Note that, for very large samples, this function can overlook small-scale spatial variations. addlearn_local applies an model aggregation/averaging technique to address this problem.
resf_vc(y, x, xconst = NULL, xgroup = NULL, weight = NULL, offset = NULL, x_nvc = FALSE, xconst_nvc = FALSE, x_sel = TRUE, x_nvc_sel = TRUE, xconst_nvc_sel = TRUE, nvc_num = 5, meig, method = "reml", penalty = "bic", nongauss = NULL, miniter=NULL, maxiter = 30, tol = 1e-30 )resf_vc(y, x, xconst = NULL, xgroup = NULL, weight = NULL, offset = NULL, x_nvc = FALSE, xconst_nvc = FALSE, x_sel = TRUE, x_nvc_sel = TRUE, xconst_nvc_sel = TRUE, nvc_num = 5, meig, method = "reml", penalty = "bic", nongauss = NULL, miniter=NULL, maxiter = 30, tol = 1e-30 )
y |
Vector of explained variables (N x 1) |
x |
Matrix of explanatory variables assuming SVC/STVC (N x K) |
xconst |
Matrix of explanatory variables assuming constant coefficients (N x K_c). Default is NULL |
xgroup |
Matrix of group IDs for modeling group-wise random effects. The IDs may be group numbers or group names (N x K_g). Default is NULL |
weight |
Vector of weights for samples (N x 1). If non-NULL, the adjusted R-squared value is evaluated for weighted explained variables. Default is NULL |
offset |
Vector of offset variables (N x 1). Available if y is count (y_type = "count" is specified in the |
x_nvc |
If TRUE, a non-linear function of x (NVC) is added on each varying coefficient on x to stablize the estimate. Default is FALSE |
xconst_nvc |
If TRUE, NVCs is added on each constant coefficient on xconst model estimate non-linear influence from xconst |
x_sel |
If TRUE, type of coefficient on x (STVC, SVC, or constant) is selected through a BIC minimization. If FALSE, S(T)VCs are assumed across x. Alternatively, x_sel can be given by column number(s) of x. For example, if x_sel = 2, the coefficient on the second explanatory variable is S(T)VC and the other coefficients are constants. The Default is TRUE |
x_nvc_sel |
If TRUE, with/without NVC on x is selected. If FALSE, NVCs are assumed across x. Alternatively, x_nvc_sel can be given by column number(s) of x. For example, if x_nvc_sel = 2, the coefficient on the second explanatory variable is NVC and the other coefficients are constants. The Default is TRUE |
xconst_nvc_sel |
If TRUE, with/without NVC on xconst is selected. If FALSE, NVCs are assumed across xconst. Alternatively, xconst_nvc_sel can be given by column number(s) of xconst. For example, if xconst_nvc_sel = 2, the coefficient on xconst[,2] becomes constant + NVC while the other coefficients become constants.The Default is TRUE |
nvc_num |
Number of natural spline basis functions to be used in NVC. Default is 5 |
meig |
Moran eigenvectors and eigenvalues. Output from |
method |
Estimation method. Restricted maximum likelihood method ("reml") and maximum likelihood method ("ml") are available. Default is "reml" |
penalty |
Penalty for model estimation and selection. "bic" for the Baysian information criterion-type penalty (N x log(K)) and "aic" for the Akaike information criterion (2K). Default is "bic" |
nongauss |
Parameter setup for modeling non-Gaussian continuous and count data. Output from |
miniter |
Minimum number of iterations. Default is NULL |
maxiter |
Maximum number of iterations. Default is 30 |
tol |
The tolerance for matrix inversion. Some errors regarding singular fit can be avoided by reducing the value, but the output can be unstable. Default is 1e-30 |
For modeling non-Gaussian data including count data, see nongauss_y.
b_vc |
Matrix of estimated spatially/spatio-temporally varying coefficients (S(T)VC + NVC) on x (N x K) |
bse_vc |
Matrix of standard errors for the varying coefficients on x (N x k) |
t_vc |
Matrix of t-values for the coefficients on x (N x K) |
p_vc |
Matrix of p-values for the coefficients on x (N x K) |
B_vc_s |
List of the estimated S(T)VCs in b_vc (= S(T)VC + NVC). The elements are the S(T)VCs (N x K), the standard errors (N x K), t-values (N x K), and p-values (N x K), respectively |
B_vc_n |
List of the estimated NVCs in b_vc (= S(T)VC + NVC). The elements are the NVCs (N x K), the standard errors (N x K), t-values (N x K), and p-values (N x K), respectively |
c |
Matrix with columns for the estimated coefficients on xconst, their standard errors, t-values, and p-values (K_c x 4). Effective if xconst_nvc = FALSE |
c_vc |
Matrix of estimated NVCs on xconst (N x K_c). Effective if xconst_nvc = TRUE |
cse_vc |
Matrix of standard errors for the NVCs on xconst (N x k_c). Effective if xconst_nvc = TRUE |
ct_vc |
Matrix of t-values for the NVCs on xconst (N x K_c). Effective if xconst_nvc = TRUE |
cp_vc |
Matrix of p-values for the NVCs on xconst (N x K_c). Effective if xconst_nvc = TRUE |
b_g |
List of K_g matrices with columns for the estimated group effects, their standard errors, and t-values |
s |
List of the variance parameters for the varying coefficient on x. The first element is a 2 x K matrix summarizing variance parameters for S(T)VC. The (1, k)-th element is the standard deviation of the k-th SVC, while the (2, k)-th element is the Moran's I value that is scaled to take a value between 0 (no spatial dependence) and 1 (strongest spatial dependence). Based on Griffith (2003), the scaled Moran'I value is interpretable as follows: 0.25-0.50:weak; 0.50-0.70:moderate; 0.70-0.90:strong; 0.90-1.00:marked. The second element of s is the vector of standard deviations of the NVCs |
s_c |
Vector of standard deviations of the NVCs on xconst |
s_g |
Vector of standard deviations of the group effects |
vc |
List indicating whether S(T)VC/NVC are removed or not during the BIC minimization. 1 indicates not removed (replaced with constant) whreas 0 indicates removed |
e |
Error statistics. When y_type="continuous", it includes residual standard error (resid_SE), adjusted conditional R2 (adjR2(cond)), restricted log-likelihood (rlogLik), Akaike information criterion (AIC), and Bayesian information criterion (BIC). rlogLik is replaced with log-likelihood (logLik) if method = "ml". resid_SE is replaced with the residual standard error for the transformed y (resid_SE_trans) if nongauss is specified. When y_type="count", the error statistics includes root mean squared error (RMSE), Gaussian likelihood approximating the model, AIC and BIC based on the likelihood, and the proportion of the null deviance explained by the model (deviance explained (%)). deviance explained, which is also used in the mgcv package, corresponds to the adjusted R2 in case of the linear regression |
pred |
Matrix of predicted values for y (pred) and their standard errors (pred_se) (N x 2). If y is transformed by specifying |
pred_quantile |
Matrix of the quantiles for the predicted values (N x 15). It is useful to evaluate uncertainty in the predictive value |
tr_par |
List of the parameter estimates for the tr_num SAL transformations. The k-th element of the list includes the four parameters for the k-th SAL transformation (see |
tr_bpar |
The estimated parameter in the Box-Cox transformation |
tr_y |
Vector of the transformed explaied variables |
resid |
Vector of residuals (N x 1) |
pdf |
Matrix whose first column consists of evenly spaced values within the value range of y and the second column consists of the estimated value of the probability density function for y if y_type in |
skew_kurt |
Skewness and kurtosis of the estimated probability density/mass function of y |
other |
List of other outputs, which are internally used |
Daisuke Murakami
Murakami, D., Yoshida, T., Seya, H., Griffith, D.A., and Yamagata, Y. (2017) A Moran coefficient-based mixed effects approach to investigate spatially varying relationships. Spatial Statistics, 19, 68-89.
Murakami, D., Kajita, M., Kajita, S. and Matsui, T. (2021) Compositionally-warped additive mixed modeling for a wide variety of non-Gaussian data. Spatial Statistics, 43, 100520.
Murakami, D., and Griffith, D.A. (2021) Balancing spatial and non-spatial variations in varying coefficient modeling: a remedy for spurious correlation. Geographical Analysis, DOI: 10.1111/gean.12310.
Murakami, D., Shirota, S., Kajita, S., and Kajita, S. (2024) Fast spatio-temporally varying coefficient modeling with reluctant interaction selection. ArXiv.
Griffith, D. A. (2003) Spatial autocorrelation and spatial filtering: gaining understanding through theory and scientific visualization. Springer Science & Business Media.
meigen, meigen_f, coef_marginal, besf_vc, addlearn_local
##################################################### ################ SVC modeling ####################### ##################################################### require(spdep) data(boston) y <- boston.c[, "CMEDV"] x <- boston.c[,c("CRIM", "AGE")] xconst <- boston.c[,c("ZN","DIS","RAD","NOX", "TAX","RM", "PTRATIO", "B")] xgroup <- boston.c[,"TOWN"] coords <- boston.c[,c("LON", "LAT")] meig <- meigen(coords=coords) # meig <- meigen_f(coords=coords) ## for large samples ##################################################### ####### Gaussian regression with SVC ################ res <- resf_vc(y=y,x=x,xconst=xconst,meig=meig ) res plot_s(res,0) # Spatially varying intercept plot_s(res,1) # 1st SVC (Not shown because the SVC is estimated constant) plot_s(res,2) # 2nd SVC #### For large samples (e.g., n > 5,000), the following #### additional learning often improves the modeling accuracy # res_adj<- addlearn_local(res) # res_adj # plot_s(res_adj,0) # plot_s(res_adj,1) # plot_s(res_adj,2) #### Group-level SVC (s_id) + random intercepts (xgroup) # meig_g <- meigen(coords, s_id=xgroup) # res2 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig_g,xgroup=xgroup) ##################################################### ####### Gaussian regression with SVC + NVC ########## # res3 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig, x_nvc =TRUE) # plot_s(res3,0) # Spatially varying intercept # plot_s(res3,1) # Spatial plot of the varying coefficient (SVC + NVC) on x[,1] # plot_s(res3,1,btype="svc")# Spatial plot of SVC in the coefficient # plot_s(res3,1,btype="nvc")# Spatial plot of NVC in the coefficient # plot_n(res3,1) # 1D plot of the NVC ##################################################### ######## Non-Gaussian regression with SVC ########### #### Model for non-Gaussian continuous data # - Probability distribution is estimated from data # ng4 <- nongauss_y( tr_num = 2 )# 2 SAL transformations to Gaussianize y # res4 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig, nongauss = ng4 ) # res4 # tr_num may be selected by comparing BIC # coef_marginal_vc(res4) # marginal effects from x (dy/dx) # plot(res4$pdf,type="l") # Estimated probability density function # res4$skew_kurt # Skew and kurtosis of the estimated PDF # res4$pred_quantile[1:2,]# predicted value by quantile #### Model for non-Gaussian and non-negative continuous data # - Probability distribution is estimated from data ## 2 SAL trans. + 1 Box-Cox trans. to Gaussianize y # ng5 <- nongauss_y( tr_num = 2, y_nonneg = TRUE ) # res5 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig, nongauss = ng5 ) # coef_marginal_vc(res5) #### Overdispersed Poisson model for count data # - y: count data #ng6 <- nongauss_y( y_type = "count" ) #res6 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig, nongauss = ng6 ) #### General model for count data # - y: count data # - Probability distribution is estimated from data #ng7 <- nongauss_y( y_type = "count", tr_num = 2 ) #res7 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig, nongauss = ng7 ) ##################################################### ################ STVC modeling ###################### ##################################################### # See \url{https://github.com/dmuraka/spmoran} #require(spData) #data(house) #dat0 <- st_as_sf(house) #dat <- data.frame(st_coordinates(dat0), dat0) #y <- log(dat[,"price"]) #x <- dat[,c("lotsize","TLA")] #xconst <- dat[,c("rooms","beds")] #byear <- house$yrbuilt #syear <- as.numeric(as.character(house$syear))#factor -> numeric #coords_z<- cbind(byear,syear) #meig <- meigen_f(coords=coords, coords_z=cbind(byear,syear),interact=TRUE) #res8 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig ) #res8 ## Varying intercept for byear <=1950 and syear==1998 #plot_s(res8,0, coords_z1_lim=c(-Inf, 1950),coords_z2_lim=1998) ## 1st STVCs which are significant at the 5 percent level, for byear <= 1950 #plot_s(res8,1, coords_z1_lim=c(-Inf, 1950), pmax=0.05) ## 2nd STVC for byear >= 1951 #plot_s(res8,2, coords_z1_lim=c(1951,Inf))##################################################### ################ SVC modeling ####################### ##################################################### require(spdep) data(boston) y <- boston.c[, "CMEDV"] x <- boston.c[,c("CRIM", "AGE")] xconst <- boston.c[,c("ZN","DIS","RAD","NOX", "TAX","RM", "PTRATIO", "B")] xgroup <- boston.c[,"TOWN"] coords <- boston.c[,c("LON", "LAT")] meig <- meigen(coords=coords) # meig <- meigen_f(coords=coords) ## for large samples ##################################################### ####### Gaussian regression with SVC ################ res <- resf_vc(y=y,x=x,xconst=xconst,meig=meig ) res plot_s(res,0) # Spatially varying intercept plot_s(res,1) # 1st SVC (Not shown because the SVC is estimated constant) plot_s(res,2) # 2nd SVC #### For large samples (e.g., n > 5,000), the following #### additional learning often improves the modeling accuracy # res_adj<- addlearn_local(res) # res_adj # plot_s(res_adj,0) # plot_s(res_adj,1) # plot_s(res_adj,2) #### Group-level SVC (s_id) + random intercepts (xgroup) # meig_g <- meigen(coords, s_id=xgroup) # res2 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig_g,xgroup=xgroup) ##################################################### ####### Gaussian regression with SVC + NVC ########## # res3 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig, x_nvc =TRUE) # plot_s(res3,0) # Spatially varying intercept # plot_s(res3,1) # Spatial plot of the varying coefficient (SVC + NVC) on x[,1] # plot_s(res3,1,btype="svc")# Spatial plot of SVC in the coefficient # plot_s(res3,1,btype="nvc")# Spatial plot of NVC in the coefficient # plot_n(res3,1) # 1D plot of the NVC ##################################################### ######## Non-Gaussian regression with SVC ########### #### Model for non-Gaussian continuous data # - Probability distribution is estimated from data # ng4 <- nongauss_y( tr_num = 2 )# 2 SAL transformations to Gaussianize y # res4 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig, nongauss = ng4 ) # res4 # tr_num may be selected by comparing BIC # coef_marginal_vc(res4) # marginal effects from x (dy/dx) # plot(res4$pdf,type="l") # Estimated probability density function # res4$skew_kurt # Skew and kurtosis of the estimated PDF # res4$pred_quantile[1:2,]# predicted value by quantile #### Model for non-Gaussian and non-negative continuous data # - Probability distribution is estimated from data ## 2 SAL trans. + 1 Box-Cox trans. to Gaussianize y # ng5 <- nongauss_y( tr_num = 2, y_nonneg = TRUE ) # res5 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig, nongauss = ng5 ) # coef_marginal_vc(res5) #### Overdispersed Poisson model for count data # - y: count data #ng6 <- nongauss_y( y_type = "count" ) #res6 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig, nongauss = ng6 ) #### General model for count data # - y: count data # - Probability distribution is estimated from data #ng7 <- nongauss_y( y_type = "count", tr_num = 2 ) #res7 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig, nongauss = ng7 ) ##################################################### ################ STVC modeling ###################### ##################################################### # See \url{https://github.com/dmuraka/spmoran} #require(spData) #data(house) #dat0 <- st_as_sf(house) #dat <- data.frame(st_coordinates(dat0), dat0) #y <- log(dat[,"price"]) #x <- dat[,c("lotsize","TLA")] #xconst <- dat[,c("rooms","beds")] #byear <- house$yrbuilt #syear <- as.numeric(as.character(house$syear))#factor -> numeric #coords_z<- cbind(byear,syear) #meig <- meigen_f(coords=coords, coords_z=cbind(byear,syear),interact=TRUE) #res8 <- resf_vc(y=y,x=x,xconst=xconst,meig=meig ) #res8 ## Varying intercept for byear <=1950 and syear==1998 #plot_s(res8,0, coords_z1_lim=c(-Inf, 1950),coords_z2_lim=1998) ## 1st STVCs which are significant at the 5 percent level, for byear <= 1950 #plot_s(res8,1, coords_z1_lim=c(-Inf, 1950), pmax=0.05) ## 2nd STVC for byear >= 1951 #plot_s(res8,2, coords_z1_lim=c(1951,Inf))
This function extracts eigenvectors and eigenvalues from a spatial weight matrix.
weigen( x = NULL, type = "knn", k = 4, threshold = 0.25, enum = NULL )weigen( x = NULL, type = "knn", k = 4, threshold = 0.25, enum = NULL )
x |
Matrix of spatial point coordinates (N x 2), sf polygon object (N spatial units), or an user-specified spatial weight matrix (N x N) (see Details) |
type |
Type of spatial weights. The currently available options are "knn" for the k-nearest neighbor-based weights, and "tri" for the Delaunay triangulation-based weights. If sf polygons are provided for x, type is ignored, and the rook-type neighborhood matrix is created |
k |
Number of nearest neighbors. It is used if type ="knn" |
threshold |
Threshold for the eigenvalues (scalar). Suppose that lambda_1 is the maximum eigenvalue. Then, this fucntion extracts eigenvectors whose corresponding eigenvalues are equal or greater than [threshold x lambda_1]. It must be a value between 0 and 1. Default is 0.25 (see Details) |
enum |
Optional. The muximum acceptable mumber of eigenvectors to be used for spatial modeling (scalar) |
If user-specified spatial weight matrix is provided for x, this function returns the eigen-pairs of the matrix. Otherwise, if sf polygon object is provided to x, the rook-type neighborhood matrix is created using this polygon, and eigen-decomposed. Otherwise, if point coordinats are provided to x, a spatial weight matrix is created according to type, and eigen-decomposed.
By default, the ARPACK routine is implemented for fast eigen-decomposition.
threshold = 0.25 (default) is a standard setting for topology-based ESF (see Tiefelsdorf and Griffith, 2007) while threshold = 0.00 is a usual setting for distance-based ESF.
sf |
Matrix of the first L eigenvectors (N x L) |
ev |
Vector of the first L eigenvalues (L x 1) |
other |
List of other outcomes, which are internally used |
Daisuke Murakami
Tiefelsdorf, M. and Griffith, D.A. (2007) Semiparametric filtering of spatial autocorrelation: the eigenvector approach. Environment and Planning A, 39 (5), 1193-1221.
Murakami, D. and Griffith, D.A. (2018) Low rank spatial econometric models. Arxiv, 1810.02956.
require(spdep) data(boston) ########## Rook adjacency-based W poly <- st_read(system.file("shapes/boston_tracts.shp",package="spData")[1]) weig1 <- weigen( poly ) ########## knn-based W coords <- boston.c[,c("LON", "LAT")] weig2 <- weigen( coords, type = "knn" ) ########## Delaunay triangulation-based W coords <- boston.c[,c("LON", "LAT")] weig3 <- weigen( coords, type = "tri") ########## User-specified W dmat <- as.matrix(dist(coords)) cmat <- exp(-dmat) diag(cmat)<- 0 weig4 <- weigen( cmat, threshold = 0 )require(spdep) data(boston) ########## Rook adjacency-based W poly <- st_read(system.file("shapes/boston_tracts.shp",package="spData")[1]) weig1 <- weigen( poly ) ########## knn-based W coords <- boston.c[,c("LON", "LAT")] weig2 <- weigen( coords, type = "knn" ) ########## Delaunay triangulation-based W coords <- boston.c[,c("LON", "LAT")] weig3 <- weigen( coords, type = "tri") ########## User-specified W dmat <- as.matrix(dist(coords)) cmat <- exp(-dmat) diag(cmat)<- 0 weig4 <- weigen( cmat, threshold = 0 )