Deploying to r-dev from @ 53156f1 🚀

Author: andrewherren

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: stochtree
 Title: Stochastic Tree Ensembles (XBART and BART) for Supervised Learning and Causal Inference
-Version: 0.2.0.9000
+Version: 0.2.1
 Authors@R: 
   c(
     person("Drew", "Herren", email = "[email protected]", role = c("aut", "cre"), comment = c(ORCID = "0000-0003-4109-6611")), 

NEWS.md

@@ -1,17 +1,13 @@
-# stochtree (development version)
-
-## New Features
-
-## Computational Improvements
+# stochtree 0.2.1
 
 ## Bug Fixes
 
-* Predict random effects correctly in R for univariate random effects models ([#248](https://github.com/StochasticTree/stochtree/pull/248))
-
-## Documentation Improvements
+* Fix prediction bug for univariate random effects models in R ([#248](https://github.com/StochasticTree/stochtree/pull/248))
 
 ## Other Changes
 
+* Encode expectations about which combinations of BART / BCF features work together and ensure warning ([#250](https://github.com/StochasticTree/stochtree/pull/250))
+
 # stochtree 0.2.0
 
 ## New Features

R/bart.R

@@ -835,6 +835,16 @@ bart <- function(
     }
   }
 
+  # Runtime checks for variance forest
+  if (include_variance_forest) {
+    if (sample_sigma2_global) {
+      warning(
+        "Global error variance will not be sampled with a heteroskedasticity forest"
+      )
+      sample_sigma2_global <- F
+    }
+  }
+
   # Handle standardization, prior calibration, and initialization of forest
   # differently for binary and continuous outcomes
   if (probit_outcome_model) {
@@ -2124,7 +2134,6 @@ predict.bartmodel <- function(
   X <- preprocessPredictionData(X, train_set_metadata)
 
   # Recode group IDs to integer vector (if passed as, for example, a vector of county names, etc...)
-  has_rfx <- FALSE
   if (predict_rfx) {
     if (!is.null(rfx_group_ids)) {
       rfx_unique_group_ids <- object$rfx_unique_group_ids
@@ -2135,7 +2144,6 @@ predict.bartmodel <- function(
         )
       }
       rfx_group_ids <- as.integer(group_ids_factor)
-      has_rfx <- TRUE
     }
   }
 

R/bcf.R

@@ -897,14 +897,7 @@ bcf <- function(
   # Handle multivariate treatment
   has_multivariate_treatment <- ncol(Z_train) > 1
   if (has_multivariate_treatment) {
-    # Disable adaptive coding, internal propensity model, and
-    # leaf scale sampling if treatment is multivariate
-    if (adaptive_coding) {
-      warning(
-        "Adaptive coding is incompatible with multivariate treatment and will be ignored"
-      )
-      adaptive_coding <- FALSE
-    }
+    # Disable internal propensity model and leaf scale sampling if treatment is multivariate
     if (is.null(propensity_train)) {
       if (propensity_covariate != "none") {
         warning(
@@ -949,21 +942,31 @@ bcf <- function(
       }
       has_basis_rfx <- TRUE
       num_basis_rfx <- ncol(rfx_basis_train)
-    } else if (rfx_model_spec == "intercept_only") {
-      rfx_basis_train <- matrix(
-        rep(1, nrow(X_train)),
-        nrow = nrow(X_train),
-        ncol = 1
-      )
-      has_basis_rfx <- TRUE
-      num_basis_rfx <- 1
     } else if (rfx_model_spec == "intercept_plus_treatment") {
-      rfx_basis_train <- cbind(
-        rep(1, nrow(X_train)),
-        Z_train
-      )
-      has_basis_rfx <- TRUE
-      num_basis_rfx <- 1 + ncol(Z_train)
+      if (has_multivariate_treatment) {
+        warning(
+          "Random effects `intercept_plus_treatment` specification is not currently implemented for multivariate treatments. This model will be fit under the `intercept_only` specification instead. Please provide a custom `rfx_basis_train` if you wish to have random slopes on multivariate treatment variables."
+        )
+        rfx_model_spec <- "intercept_only"
+      }
+    }
+    if (is.null(rfx_basis_train)) {
+      if (rfx_model_spec == "intercept_only") {
+        rfx_basis_train <- matrix(
+          rep(1, nrow(X_train)),
+          nrow = nrow(X_train),
+          ncol = 1
+        )
+        has_basis_rfx <- TRUE
+        num_basis_rfx <- 1
+      } else {
+        rfx_basis_train <- cbind(
+          rep(1, nrow(X_train)),
+          Z_train
+        )
+        has_basis_rfx <- TRUE
+        num_basis_rfx <- 1 + ncol(Z_train)
+      }
     }
     num_rfx_groups <- length(unique(rfx_group_ids_train))
     num_rfx_components <- ncol(rfx_basis_train)
@@ -1021,15 +1024,21 @@ bcf <- function(
     y_train <- as.matrix(y_train)
   }
 
-  # Check whether treatment is binary (specifically 0-1 binary)
-  binary_treatment <- length(unique(Z_train)) == 2
-  if (binary_treatment) {
-    unique_treatments <- sort(unique(Z_train))
-    if (!(all(unique_treatments == c(0, 1)))) binary_treatment <- FALSE
+  # Check whether treatment is binary and univariate (specifically 0-1 binary)
+  binary_treatment <- FALSE
+  if (!has_multivariate_treatment) {
+    binary_treatment <- length(unique(Z_train)) == 2
+    if (binary_treatment) {
+      unique_treatments <- sort(unique(Z_train))
+      if (!(all(unique_treatments == c(0, 1)))) binary_treatment <- FALSE
+    }
   }
 
   # Adaptive coding will be ignored for continuous / ordered categorical treatments
   if ((!binary_treatment) && (adaptive_coding)) {
+    warning(
+      "Adaptive coding is only compatible with binary (univariate) treatment and, as a result, will be ignored in sampling this model"
+    )
     adaptive_coding <- FALSE
   }
 

R/posterior_transformation.R

@@ -409,19 +409,31 @@ compute_contrast_bart_model <- function(
       "rfx_group_ids_0 and rfx_group_ids_1 must be provided for this model"
     )
   }
-  if ((has_rfx) && (is.null(rfx_basis_0) || is.null(rfx_basis_1))) {
-    stop(
-      "rfx_basis_0 and rfx_basis_1 must be provided for this model"
-    )
-  }
-  if (
-    (object$model_params$num_rfx_basis > 0) &&
-      ((ncol(rfx_basis_0) != object$model_params$num_rfx_basis) ||
-        (ncol(rfx_basis_1) != object$model_params$num_rfx_basis))
-  ) {
-    stop(
-      "rfx_basis_0 and / or rfx_basis_1 have a different dimension than the basis used to train this model"
-    )
+  if (has_rfx) {
+    if (object$model_params$rfx_model_spec == "custom") {
+      if ((is.null(rfx_basis_0) || is.null(rfx_basis_1))) {
+        stop(
+          "A user-provided basis (`rfx_basis_0` and `rfx_basis_1`) must be provided when the model was sampled with a random effects model spec set to 'custom'"
+        )
+      }
+      if (!is.matrix(rfx_basis_0) || !is.matrix(rfx_basis_1)) {
+        stop("'rfx_basis_0' and 'rfx_basis_1' must be matrices")
+      }
+      if ((nrow(rfx_basis_0) != nrow(X)) || (nrow(rfx_basis_1) != nrow(X))) {
+        stop(
+          "'rfx_basis_0' and 'rfx_basis_1' must have the same number of rows as 'X'"
+        )
+      }
+      if (
+        (object$model_params$num_rfx_basis > 0) &&
+          ((ncol(rfx_basis_0) != object$model_params$num_rfx_basis) ||
+            (ncol(rfx_basis_1) != object$model_params$num_rfx_basis))
+      ) {
+        stop(
+          "rfx_basis_0 and / or rfx_basis_1 have a different dimension than the basis used to train this model"
+        )
+      }
+    }
   }
 
   # Predict for the control arm
@@ -574,16 +586,22 @@ sample_bcf_posterior_predictive <- function(
         "'rfx_group_ids' must have the same length as the number of rows in 'X'"
       )
     }
-    if (is.null(rfx_basis)) {
-      stop(
-        "'rfx_basis' must be provided in order to compute the requested intervals"
-      )
-    }
-    if (!is.matrix(rfx_basis)) {
-      stop("'rfx_basis' must be a matrix")
+
+    if (model_object$model_params$rfx_model_spec == "custom") {
+      if (is.null(rfx_basis)) {
+        stop(
+          "A user-provided basis (`rfx_basis`) must be provided when the model was sampled with a random effects model spec set to 'custom'"
+        )
+      }
     }
-    if (nrow(rfx_basis) != nrow(X)) {
-      stop("'rfx_basis' must have the same number of rows as 'X'")
+
+    if (!is.null(rfx_basis)) {
+      if (!is.matrix(rfx_basis)) {
+        stop("'rfx_basis' must be a matrix")
+      }
+      if (nrow(rfx_basis) != nrow(X)) {
+        stop("'rfx_basis' must have the same number of rows as 'X'")
+      }
     }
   }
 
@@ -735,16 +753,18 @@ sample_bart_posterior_predictive <- function(
         "'rfx_group_ids' must have the same length as the number of rows in 'X'"
       )
     }
-    if (is.null(rfx_basis)) {
-      stop(
-        "'rfx_basis' must be provided in order to compute the requested intervals"
-      )
-    }
-    if (!is.matrix(rfx_basis)) {
-      stop("'rfx_basis' must be a matrix")
-    }
-    if (nrow(rfx_basis) != nrow(X)) {
-      stop("'rfx_basis' must have the same number of rows as 'X'")
+    if (model_object$model_params$rfx_model_spec == "custom") {
+      if (is.null(rfx_basis)) {
+        stop(
+          "A user-provided basis (`rfx_basis`) must be provided when the model was sampled with a random effects model spec set to 'custom'"
+        )
+      }
+      if (!is.matrix(rfx_basis)) {
+        stop("'rfx_basis' must be a matrix")
+      }
+      if (nrow(rfx_basis) != nrow(X)) {
+        stop("'rfx_basis' must have the same number of rows as 'X'")
+      }
     }
   }
 
@@ -1172,16 +1192,18 @@ compute_bart_posterior_interval <- function(
         "'rfx_group_ids' must have the same length as the number of rows in 'X'"
       )
     }
-    if (is.null(rfx_basis)) {
-      stop(
-        "'rfx_basis' must be provided in order to compute the requested intervals"
-      )
-    }
-    if (!is.matrix(rfx_basis)) {
-      stop("'rfx_basis' must be a matrix")
-    }
-    if (nrow(rfx_basis) != nrow(X)) {
-      stop("'rfx_basis' must have the same number of rows as 'X'")
+    if (model_object$model_params$rfx_model_spec == "custom") {
+      if (is.null(rfx_basis)) {
+        stop(
+          "A user-provided basis (`rfx_basis`) must be provided when the model was sampled with a random effects model spec set to 'custom'"
+        )
+      }
+      if (!is.matrix(rfx_basis)) {
+        stop("'rfx_basis' must be a matrix")
+      }
+      if (nrow(rfx_basis) != nrow(X)) {
+        stop("'rfx_basis' must have the same number of rows as 'X'")
+      }
     }
   }
 

configure

@@ -1,6 +1,6 @@
 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.72 for stochtree 0.2.0.9000.
+# Generated by GNU Autoconf 2.72 for stochtree 0.2.1.
 #
 #
 # Copyright (C) 1992-1996, 1998-2017, 2020-2023 Free Software Foundation,
@@ -600,8 +600,8 @@ MAKEFLAGS=
 # Identity of this package.
 PACKAGE_NAME='stochtree'
 PACKAGE_TARNAME='stochtree'
-PACKAGE_VERSION='0.2.0.9000'
-PACKAGE_STRING='stochtree 0.2.0.9000'
+PACKAGE_VERSION='0.2.1'
+PACKAGE_STRING='stochtree 0.2.1'
 PACKAGE_BUGREPORT=''
 PACKAGE_URL=''
 
@@ -1204,7 +1204,7 @@ if test "$ac_init_help" = "long"; then
   # Omit some internal or obsolete options to make the list less imposing.
   # This message is too long to be a string in the A/UX 3.1 sh.
   cat <<_ACEOF
-'configure' configures stochtree 0.2.0.9000 to adapt to many kinds of systems.
+'configure' configures stochtree 0.2.1 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1266,7 +1266,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of stochtree 0.2.0.9000:";;
+     short | recursive ) echo "Configuration of stochtree 0.2.1:";;
    esac
   cat <<\_ACEOF
 
@@ -1334,7 +1334,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-stochtree configure 0.2.0.9000
+stochtree configure 0.2.1
 generated by GNU Autoconf 2.72
 
 Copyright (C) 2023 Free Software Foundation, Inc.
@@ -1371,7 +1371,7 @@ cat >config.log <<_ACEOF
 This file contains any messages produced by compilers while
 running configure, to aid debugging if configure makes a mistake.
 
-It was created by stochtree $as_me 0.2.0.9000, which was
+It was created by stochtree $as_me 0.2.1, which was
 generated by GNU Autoconf 2.72.  Invocation command line was
 
   $ $0$ac_configure_args_raw
@@ -2380,7 +2380,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
 # report actual input values of CONFIG_FILES etc. instead of their
 # values after options handling.
 ac_log="
-This file was extended by stochtree $as_me 0.2.0.9000, which was
+This file was extended by stochtree $as_me 0.2.1, which was
 generated by GNU Autoconf 2.72.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -2435,7 +2435,7 @@ ac_cs_config_escaped=`printf "%s\n" "$ac_cs_config" | sed "s/^ //; s/'/'\\\\\\\\
 cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
 ac_cs_config='$ac_cs_config_escaped'
 ac_cs_version="\\
-stochtree config.status 0.2.0.9000
+stochtree config.status 0.2.1
 configured by $0, generated by GNU Autoconf 2.72,
   with options \\"\$ac_cs_config\\"
 

configure.ac

@@ -3,7 +3,7 @@
 # https://github.com/microsoft/LightGBM/blob/master/R-package/configure.ac
 
 AC_PREREQ(2.69)
-AC_INIT([stochtree], [0.2.0.9000], [], [stochtree], [])
+AC_INIT([stochtree], [0.2.1], [], [stochtree], [])
 # Note: consider making version number dynamic as in 
 # https://github.com/microsoft/LightGBM/blob/195c26fc7b00eb0fec252dfe841e2e66d6833954/build-cran-package.sh