Sciome Update 1/10/2024

.github/workflows/check-standard.yaml

@@ -18,11 +18,8 @@ jobs:
       fail-fast: false
       matrix:
         config:
-          - {os: macos-latest,   r: 'release'}
-          - {os: windows-latest, r: 'release'}
-          - {os: ubuntu-latest,   r: 'devel', http-user-agent: 'release'}
           - {os: ubuntu-latest,   r: 'release'}
-          - {os: ubuntu-latest,   r: 'oldrel-1'}
+
 
     env:
       GITHUB_PAT: ${{ secrets.GITHUB_TOKEN }}
@@ -41,7 +38,12 @@ jobs:
 
       - uses: r-lib/actions/setup-r-dependencies@v2
         with:
-          extra-packages: any::rcmdcheck
+          extra-packages: |
+            any::tictoc
+            any::units
+            any::rcmdcheck
+            any::knitr
+            any::rmarkdown
           needs: check
 
       - uses: r-lib/actions/check-r-package@v2

DESCRIPTION

@@ -1,7 +1,7 @@
 Package: PrestoGP
 Type: Package
 Title: Penalized Regression for Spatio-Temporal Outcomes via Gaussian Processes 
-Version: 0.2.0.9020
+Version: 0.2.0.9021
 Authors@R: c(
     person(given = "Eric",
            family = "Bair",
@@ -45,7 +45,9 @@ Imports:
     rlang,
     mvtnorm,
     spam,
-    psych
+    psych,
+    doParallel,
+    covr
 License: GPL-3
 Encoding: UTF-8
 VignetteBuilder: knitr 
@@ -55,10 +57,8 @@ Collate:
     'PrestoGP-package.R'
     'PrestoGP_CreateU_Multivariate.R'
     'PrestoGP_Model.R'
-    'PrestoGP_Vecchia_Spatiotemporal.R'
+    'PrestoGP_Vecchia.R'
     'PrestoGP_Full.R'
-    'PrestoGP_Vecchia_Spatial.R'
-    'PrestoGP_Full_Spatial.R'
     'PrestoGP_Multivariate_Vecchia.R'
     'PrestoGP_Util_Functions.R'
     'RcppExports.R'

NAMESPACE

@@ -2,14 +2,12 @@
 
 export(":=")
 export(.data)
-export(FullSpatialModel)
+export(FullModel)
 export(Kr_pred)
 export(MultivariateVecchiaModel)
 export(PrestoGPModel)
 export(ST_Krig_Param_Avg)
-export(SpatialModel)
-export(SpatiotemporalFullModel)
-export(SpatiotemporalModel)
+export(VecchiaModel)
 export(as_label)
 export(as_name)
 export(createUMultivariate)
@@ -31,6 +29,7 @@ import(glmnet)
 import(ncvreg)
 import(readxl)
 import(scoringRules)
+import(covr)
 importFrom(aod,wald.test)
 importFrom(dplyr,"%>%")
 importFrom(foreach,"%dopar%")

R/Log_Likelihood.R

@@ -12,11 +12,18 @@
 #'
 #' @examples
 #' @noRd
-negloglik_vecchia_ST <- function(logparms, locs, res, vecchia.approx) {
-  parms <- exp(logparms)
-  locs <- locs / matrix(parms[c(2, 2, 3)], nrow = nrow(locs), ncol = 3, byrow = TRUE)
-  vecchia.approx$locsord <- locs
-  -vecchia_likelihood(res, vecchia.approx, c(parms[1], 1, 0.5), parms[4])
+negloglik_vecchia_ST <- function(logparms, res, vecchia.approx, param.seq,
+                                 scaling, nscale) {
+  parms <- unlog.params(logparms, param.seq, 1)
+  locs.scaled <- vecchia.approx$locsord
+  for (j in 1:nscale) {
+    locs.scaled[, scaling == j] <- locs.scaled[, scaling == j] /
+      parms[param.seq[2, 1] + j - 1]
+  }
+  vecchia.approx$locsord <- locs.scaled
+  -vecchia_likelihood(res, vecchia.approx, c(parms[1], 1,
+                                             parms[param.seq[3, 1]]),
+                      parms[param.seq[4, 1]])
 }
 
 #' negloglik_vecchia
@@ -33,9 +40,10 @@
 #'
 #' @examples
 #' @noRd
-negloglik_vecchia <- function(logparms, locs, res, vecchia.approx) {
-  parms <- exp(logparms)
-  -vecchia_likelihood(res, vecchia.approx, c(parms[1], parms[2], 0.5), parms[3])
+negloglik_vecchia <- function(logparms, res, vecchia.approx, param.seq) {
+  parms <- unlog.params(logparms, param.seq, 1)
+  -vecchia_likelihood(res, vecchia.approx, c(parms[1], parms[2], parms[3]),
+                      parms[4])
 }
 
 #' negloglik_full_ST
@@ -52,12 +60,21 @@
 #'
 #' @examples
 #' @noRd
-negloglik_full_ST <- function(logparms, locs, y, N) {
-  parms <- exp(logparms)
-  locs.scaled <- cbind(locs[, 1] / parms[2], locs[, 2] / parms[2], locs[, 3] / parms[3])
+negloglik_full_ST <- function(logparms, locs, y, param.seq, scaling, nscale) {
+  parms <- unlog.params(logparms, param.seq, 1)
+  locs.scaled <- locs
+  for (j in 1:nscale) {
+    locs.scaled[, scaling == j] <- locs.scaled[, scaling == j] /
+      parms[param.seq[2, 1] + j - 1]
+  }
   d <- fields::rdist(locs.scaled)
-  cov.mat <- parms[1] * fields::Exponential(d, range = 1) + parms[4] * diag(N)
-  -mvtnorm::dmvnorm(y, rep(0, N), cov.mat, log = TRUE)
+  N <- nrow(d)
+  cov.mat <- parms[1] * fields::Matern(d,
+    range = 1,
+    smoothness = parms[param.seq[3, 1]]
+  ) +
+    parms[param.seq[4, 1]] * diag(N)
+  return(-1 * mvtnorm::dmvnorm(y, rep(0, N), cov.mat, log = TRUE))
 }
 
 #' negloglik.full
@@ -74,15 +91,12 @@
 #'
 #' @examples
 #' @noRd
-negloglik.full <- function(logparams, locs, y) {
-  params <- c(
-    exp(logparams[1:2]),
-    gtools::inv.logit(logparams[3], 0, 2.5),
-    exp(logparams[4])
-  )
-  d <- fields::rdist(locs)
+negloglik.full <- function(logparams, d, y, param.seq) {
+  params <- unlog.params(logparams, param.seq, 1)
+  #    d <- fields::rdist(locs)
   N <- nrow(d)
-  cov.mat <- params[1] * fields::Matern(d, range = params[2], smoothness = params[3]) +
+  cov.mat <- params[1] * fields::Matern(d, range = params[2],
+                                        smoothness = params[3]) +
     params[4] * diag(N)
   return(-1 * mvtnorm::dmvnorm(y, rep(0, N), cov.mat, log = TRUE))
 }
@@ -94,10 +108,10 @@
 mvnegloglik <- function(logparams, vecchia.approx, y, param.seq, P) {
   #  Input-
   #  logparams: A numeric vector of length (4*P)+(4*choose(P,2)).
   #             To construct these parameters we unlist a list of the 7 covariance
   #             categories- in order: (1) marginal variances, (2) Marginal ranges,
   #             (3) Marginal smoothness, (4) Nuggets, and
   #             (5) cross-covariance correlation. These seven parameters are to be
   #             created in a list. The variance, range, smoothness, and nugget
   #             have P terms, and the correlations have choose(P,2)
   #             terms. Use unlist() to create the vector of parameters.
@@ -110,32 +124,23 @@
   #           index locations of each parameter.
 
   # P <- length(y)
   # transform the postively constrained parameters from log-space to normal-space
-  params <- c(
-    exp(logparams[1:param.seq[2, 2]]),
-    gtools::inv.logit(logparams[param.seq[3, 1]:param.seq[3, 2]], 0, 2.5),
-    exp(logparams[param.seq[4, 1]:param.seq[4, 2]])
-  )
-  if (P > 1) {
-    params <- c(params, tanh(logparams[param.seq[5, 1]:param.seq[5, 2]]))
-  } else {
-    params <- c(params, 1)
-  }
-
+  params <- unlog.params(logparams, param.seq, P)
   U.obj <- createUMultivariate(vecchia.approx, params)
   -1 * GPvecchia:::vecchia_likelihood_U(y, U.obj)
 }
 
 ##############################################################################
 ### Flexible Spatiotemporal Multivariate Matern Negative Loglikelihood Function ###########
 
-mvnegloglik_ST <- function(logparams, vecchia.approx, y, param.seq, P, scaling, nscale) {
+mvnegloglik_ST <- function(logparams, vecchia.approx, y, param.seq, P, scaling,
+                           nscale) {
   #  Input-
   #  logparams: A numeric vector of length (4*P)+(4*choose(P,2)).
   #             To construct these parameters we unlist a list of the 7 covariance
   #             categories- in order: (1) marginal variances, (2) Marginal ranges,
   #             (3) Marginal smoothness, (4) Nuggets, and
   #             (5) cross-covariance correlation. These seven parameters are to be
   #             created in a list. The variance, range, smoothness, and nugget
   #             have P terms, and the correlations have choose(P,2)
   #             terms. Use unlist() to create the vector of parameters.
@@ -148,17 +153,8 @@
   #           index locations of each parameter.
 
   # P <- length(y)
   # transform the postively constrained parameters from log-space to normal-space
-  params <- c(
-    exp(logparams[1:param.seq[2, 2]]),
-    gtools::inv.logit(logparams[param.seq[3, 1]:param.seq[3, 2]], 0, 2.5),
-    exp(logparams[param.seq[4, 1]:param.seq[4, 2]])
-  )
-  if (P > 1) {
-    params <- c(params, tanh(logparams[param.seq[5, 1]:param.seq[5, 2]]))
-  } else {
-    params <- c(params, 1)
-  }
+  params <- unlog.params(logparams, param.seq, P)
   locs.scaled <- vecchia.approx$locsord
   for (i in 1:P) {
     for (j in 1:nscale) {
@@ -201,15 +197,7 @@
   # P <- length(y)
   # transform the postively constrained parameters from log-space to normal-space
   P <- length(locs)
-  params <- c(
-    exp(logparams[1:param.seq[2, 2]]),
-    gtools::inv.logit(logparams[param.seq[3, 1]:param.seq[3, 2]], 0, 2.5),
-    exp(logparams[param.seq[4, 1]:param.seq[4, 2]])
-  )
-  if (P > 1) {
-    params <- c(params, tanh(logparams[param.seq[5, 1]:param.seq[5, 2]]))
-  }
-
+  params <- unlog.params(logparams, param.seq, P)
   sig2 <- params[param.seq[1, 1]:param.seq[1, 2]]
   range <- params[param.seq[2, 1]:param.seq[2, 2]]
   smoothness <- params[param.seq[3, 1]:param.seq[3, 2]]
@@ -254,11 +242,15 @@
       j <- combs[iter, 2]
 
       smoothness.mat[i, j] <- (marg.smooth[i] + marg.smooth[j]) / 2
-      range.mat[i, j] <- 1 / sqrt(((1 / marg.range[i])^2 + (1 / marg.range[j])^2) / 2)
+      range.mat[i, j] <- 1 / sqrt(((1 / marg.range[i])^2 +
+                                     (1 / marg.range[j])^2) / 2)
 
       s1 <- sqrt(marg.var[i] * marg.var[j])
-      s2 <- ((1 / marg.range[i])^marg.smooth[i] * (1 / marg.range[j])^marg.smooth[j]) / ((1 / range.mat[i, j])^(2 * smoothness.mat[i, j]))
-      s3 <- gamma(smoothness.mat[i, j]) / (sqrt(gamma(marg.smooth[i])) * sqrt(gamma(marg.smooth[j])))
+      s2 <- ((1 / marg.range[i])^marg.smooth[i] *
+               (1 / marg.range[j])^marg.smooth[j]) /
+        ((1 / range.mat[i, j])^(2 * smoothness.mat[i, j]))
+      s3 <- gamma(smoothness.mat[i, j]) / (sqrt(gamma(marg.smooth[i])) *
+                                             sqrt(gamma(marg.smooth[j])))
       s4 <- R.corr[iter]
       sig2.mat[i, j] <- s1 * s2 * s3 * s4
     }
@@ -300,10 +292,12 @@
     # Calculate the covariance matrix - if/then based on its location in the super-matrix
     N <- nrow(d)
     if (i == j) { # To accomodate varying size outcomes- the nugget is not included on cross-covariances
-      cov.mat.ij <- sig2[i, j] * geoR::matern(d, phi = range[i, j], kappa = smoothness[i, j]) +
+      cov.mat.ij <- sig2[i, j] * geoR::matern(d, phi = range[i, j], kappa =
+                                                smoothness[i, j]) +
         nugget[i, j] * diag(N)
     } else {
-      cov.mat.ij <- sig2[i, j] * geoR::matern(d, phi = range[i, j], kappa = smoothness[i, j])
+      cov.mat.ij <- sig2[i, j] * geoR::matern(d, phi = range[i, j], kappa =
+                                                smoothness[i, j])
     }
 
 
@@ -349,3 +343,20 @@
   }
   return(logparams.init)
 }
+
+##############################################################################
+### Transform the log Matern parameters back to the original #########
+
+unlog.params <- function(logparams, param.seq, P) {
+  params <- c(
+    exp(logparams[1:param.seq[2, 2]]),
+    gtools::inv.logit(logparams[param.seq[3, 1]:param.seq[3, 2]], 0, 2.5),
+    exp(logparams[param.seq[4, 1]:param.seq[4, 2]])
+  )
+  if (P > 1) {
+    params <- c(params, tanh(logparams[param.seq[5, 1]:param.seq[5, 2]]))
+  } else {
+    params <- c(params, 1)
+  }
+  return(params)
+}

R/PrestoGP_CreateU_Multivariate.R

@@ -66,7 +66,8 @@ max_min_ordering <- function(locs, dist_func) {
 #' @param dist_func Any distance function with a signature of dist(query_location, locations_matrix)
 #'
 #' @return A vector containing the indices of the neighbors
-knn_indices <- function(ordered_locs, query, n_neighbors, dist_func, dist_func_code) {
+knn_indices <- function(ordered_locs, query, n_neighbors,
+                        dist_func, dist_func_code) {
   if (dist_func_code == "custom") {
     dists <- dist_func(query, ordered_locs)
     dists_order <- order(dists)
@@ -90,25 +91,34 @@ knn_indices <- function(ordered_locs, query, n_neighbors, dist_func, dist_func_c
 #' @param dist_func Any distance function with a signature of dist(query_location, locations_matrix)
 #'
 #' @return A list containing two matrices, each with one row per location: an indices matrix with the indices of nearest neighbors for each location, and a distance matrix with the associated distances
-sparseNN <- function(ordered_locs, n_neighbors, dist_func, dist_func_code, ordered_locs_pred = NULL) {
+sparseNN <- function(ordered_locs, n_neighbors,
+                     dist_func, dist_func_code, ordered_locs_pred = NULL) {
   ee <- min(apply(ordered_locs, 2, stats::sd))
   n <- nrow(ordered_locs)
   ordered_locs <- ordered_locs + matrix(
     ee * 1e-04 *
       stats::rnorm(n * ncol(ordered_locs)),
     n, ncol(ordered_locs)
   )
-  indices_matrix <- matrix(data = NA, nrow = nrow(ordered_locs), ncol = n_neighbors)
-  distances_matrix <- matrix(data = NA, nrow = nrow(ordered_locs), ncol = n_neighbors)
+  indices_matrix <- matrix(data = NA, nrow = nrow(ordered_locs),
+                           ncol = n_neighbors)
+  distances_matrix <- matrix(data = NA, nrow = nrow(ordered_locs),
+                             ncol = n_neighbors)
   for (row in 1:n_neighbors) {
     # for the locations from 1 to n_neighbors, use the entire locs list to find the neighbors
-    nn <- knn_indices(ordered_locs[1:(n_neighbors + 1), , drop = FALSE][-row, , drop = FALSE], ordered_locs[row, , drop = FALSE], n_neighbors, dist_func, dist_func_code)
+    nn <- knn_indices(ordered_locs[1:
+                                     (n_neighbors + 1), , drop = FALSE][-row, ,
+                                                              drop = FALSE],
+                      ordered_locs[row, , drop = FALSE], n_neighbors,
+                      dist_func, dist_func_code)
     indices_matrix[row, 1:n_neighbors] <- nn$indices[1:n_neighbors]
     distances_matrix[row, 1:n_neighbors] <- nn$distances[1:n_neighbors]
   }
   for (row in (n_neighbors + 1):nrow(ordered_locs)) {
     # get the m nearest neighbors from the locs before this one in the max-min order
-    nn <- knn_indices(ordered_locs[1:(row - 1), , drop = FALSE], ordered_locs[row, , drop = FALSE], n_neighbors, dist_func, dist_func_code)
+    nn <- knn_indices(ordered_locs[1:(row - 1), , drop = FALSE],
+                      ordered_locs[row, , drop = FALSE], n_neighbors,
+                      dist_func, dist_func_code)
     indices_matrix[row, 1:n_neighbors] <- nn$indices[1:n_neighbors]
     distances_matrix[row, 1:n_neighbors] <- nn$distances[1:n_neighbors]
   }