Debian Codegraph @

R/gls.R

raw
1 #-------------------------------------
2 # vcovCR with defaults
3 #-------------------------------------
4
5 #' Cluster-robust variance-covariance matrix for a gls object.
6 #'
7 #' \code{vcovCR} returns a sandwich estimate of the variance-covariance matrix
8 #' of a set of regression coefficient estimates from a \code{\link[nlme]{gls}} object.
9 #'
10 #' @param cluster Optional expression or vector indicating which observations
11 #' belong to the same cluster. If not specified, will be set to
12 #' \code{getGroups(obj)}.
13 #' @param target Optional matrix or vector describing the working
14 #' variance-covariance model used to calculate the \code{CR2} and \code{CR4}
15 #' adjustment matrices. If not specified, the target is taken to be the
16 #' estimated variance-covariance structure of the \code{gls} object.
17 #' @inheritParams vcovCR
18 #'
19 #' @return An object of class \code{c("vcovCR","clubSandwich")}, which consists
20 #' of a matrix of the estimated variance of and covariances between the
21 #' regression coefficient estimates.
22 #'
23 #' @seealso \code{\link{vcovCR}}
24 #'
25 #' @examples
26 #'
27 #' if (requireNamespace("nlme", quietly = TRUE)) {
28 #'
29 #' library(nlme)
30 #' data(Ovary, package = "nlme")
31 #' Ovary$time_int <- 1:nrow(Ovary)
32 #' lm_AR1 <- gls(follicles ~ sin(2*pi*Time) + cos(2*pi*Time), data = Ovary,
33 #' correlation = corAR1(form = ~ time_int | Mare))
34 #' vcovCR(lm_AR1, type = "CR2")
35 #'
36 #' }
37 #'
38 #' @export
39
40 vcovCR.gls <- function(obj, cluster, type, target, inverse_var, form = "sandwich", ...) {
41 if (missing(cluster)) cluster <- nlme::getGroups(obj)
42 if (missing(target)) target <- NULL
43 if (missing(inverse_var) ) inverse_var <- is.null(target)
44 vcov_CR(obj, cluster = cluster, type = type,
45 target = target, inverse_var = inverse_var, form = form)
46 }
47
48 # residuals_CS()
49 # coef()
50 # nobs()
51
52 #-------------------------------------
53 # model_matrix()
54 #-------------------------------------
55
56 get_data <- function (object) {
57
58 if ("data" %in% names(object)) {
59 data <- object$data
60 } else {
61 dat_call <- object$call$data
62 envir_names <- sys.frames()
63 data <- simpleError("start")
64 i <- 1L
65 while (inherits(data, "error") & i <= length(envir_names)) {
66 data <- tryCatch(eval(dat_call, envir = envir_names[[i]]), error = function(e) e)
67 i <- i + 1L
68 }
69 }
70
71 if (inherits(data, "error")) return(NULL)
72
73 naAct <- object[["na.action"]]
74 if (!is.null(naAct)) {
75 data <- if (inherits(naAct, "omit")) {
76 data[-naAct, ]
77
78 } else if (inherits(naAct, "exclude")) {
79 data
80 } else eval(object$call$na.action)(data)
81 }
82
83 subset <- object$call$subset
84 if (!is.null(subset)) {
85 subset <- eval(asOneSidedFormula(subset)[[2]], data)
86 data <- data[subset, ]
87 }
88
89 data
90 }
91
92 #' @export
93
94 model_matrix.gls <- function(obj) {
95 dat <- get_data(obj)
96 model.matrix(formula(obj), data = dat)
97 }
98
99 #-------------------------------------
100 # Get (model-based) working variance matrix
101 #-------------------------------------
102
103 #' @export
104
105 targetVariance.gls <- function(obj, cluster = nlme::getGroups(obj)) {
106
107 groups <- nlme::getGroups(obj)
108 if (is.null(groups)) groups <- cluster
109
110 if (is.null(obj$modelStruct$corStruct)) {
111 if (is.null(obj$modelStruct$varStruct)) {
112 V_list <- matrix_list(rep(1, length(cluster)), cluster, "both")
113 } else {
114 wts <- nlme::varWeights(obj$modelStruct$varStruct)
115 V_list <- matrix_list(1 / wts^2, cluster, "both")
116 }
117 } else {
118 R_list <- nlme::corMatrix(obj$modelStruct$corStruct)
119 if (is.null(obj$modelStruct$varStruct)) {
120 V_list <- R_list
121 } else {
122 sd_vec <- 1 / nlme::varWeights(obj$modelStruct$varStruct)[order(order(groups))]
123 sd_list <- split(sd_vec, groups)
124 V_list <- Map(function(R, s) tcrossprod(s) * R, R = R_list, s = sd_list)
125 }
126 }
127
128 # check if clustering level is higher than highest level of random effects
129
130 tb_groups <- table(groups)
131 tb_cluster <- table(cluster)
132 if (length(tb_groups) < length(tb_cluster) |
133 any(as.vector(tb_groups) != rep(as.vector(tb_cluster), length.out = length(tb_groups))) |
134 any(names(tb_groups) != rep(names(tb_cluster), length.out = length(tb_groups)))) {
135
136 # check that random effects are nested within clusters
137 tb_cross <- table(groups, cluster)
138 nested <- apply(tb_cross, 1, function(x) sum(x > 0) == 1)
139 if (!all(nested)) stop("Random effects are not nested within clustering variable.")
140
141 # expand target_list to level of clustering
142 crosswalk <- data.frame(groups, cluster)
143 V_list <- add_bdiag(small_mats = V_list,
144 big_mats = matrix_list(rep(0, length(cluster)), cluster, dim = "both"),
145 crosswalk = crosswalk)
146 }
147
148 V_list
149 }
150
151 #-------------------------------------
152 # Get weighting matrix
153 #-------------------------------------
154
155 #' @export
156
157 weightMatrix.gls <- function(obj, cluster = nlme::getGroups(obj)) {
158 V_list <- targetVariance(obj, cluster)
159 lapply(V_list, function(v) chol2inv(chol(v)))
160 }
161
162 #---------------------------------------
163 # Get bread matrix and scaling constant
164 #---------------------------------------
165
166 #' @export
167
168 bread.gls <- function(x, ...) {
169 vcov(x) * nobs(x) / x$sigma^2
170 }
171
172 # v_scale() is default