v0.9.6

Author: pachadotdev

DESCRIPTION

@@ -2,7 +2,7 @@ Package: capybara
 Type: Package
 Title: Fast and Memory Efficient Fitting of Linear Models with High-Dimensional
     Fixed Effects
-Version: 0.9.4
+Version: 0.9.6
 Authors@R: c(
     person(
         given = "Mauricio",

NEWS.md

@@ -1,3 +1,15 @@
+# capybara 0.9.6
+
+* Calculates the rank of matrix X based on singular value decomposition instead
+  of QR decomposition. This is more efficient and numerically stable.
+
+# capybara 0.9.5
+
+* Fixes and expands the 'weights' argument in the `fe*()` functions to allow for
+  different types of weights. The default is still `NULL` (i.e., all weights
+  equal to 1). The argument now admits weights passed as `weights = ~cyl`,
+  `weights = mtcars$cyl`, or `w <- mtcars$cyl; weights = w`.
+
 # capybara 0.9.4
 
 * Allows to estimate models without fixed effects.

R/cpp11.R

@@ -1,5 +1,9 @@
 # Generated by cpp11: do not edit by hand
 
+check_linear_dependence_svd_ <- function(y, x, p) {
+  .Call(`_capybara_check_linear_dependence_svd_`, y, x, p)
+}
+
 center_variables_r_ <- function(V_r, w_r, klist, tol, max_iter, iter_interrupt, iter_ssr) {
   .Call(`_capybara_center_variables_r_`, V_r, w_r, klist, tol, max_iter, iter_interrupt, iter_ssr)
 }

R/feglm.R

@@ -197,13 +197,22 @@ feglm <- function(
   model_response_(data, formula)
 
   # Check for linear dependence ----
-  # check_linear_dependence_(x, p)
-  check_linear_dependence_(cbind(y, x), p + 1L)
+  check_linear_dependence_(y, x, p + 1L)
 
   # Extract weights if required ----
   if (is.null(weights)) {
     wt <- rep(1.0, nt)
+  } else if (exists("weights_vec")) {
+    # Weights provided as vector
+    wt <- weights_vec
+    if (length(wt) != nrow(data)) {
+      stop("Length of weights vector must equal number of observations.", call. = FALSE)
+    }
+  } else if (exists("weights_col")) {
+    # Weights provided as formula - use the extracted column name
+    wt <- data[[weights_col]]
   } else {
+    # Weights provided as column name
     wt <- data[[weights]]
   }
 

R/feglm_helpers.R

@@ -240,13 +240,37 @@ col_types <- function(data) {
 #' @noRd
 model_frame_ <- function(data, formula, weights) {
   # Necessary columns
-  needed_cols <- c(all.vars(formula), weights)
+  formula_vars <- all.vars(formula)
+
+  # Handle different ways weights might be specified
+  if (is.null(weights)) {
+    # No weights specified
+    weight_col <- NULL
+    needed_cols <- formula_vars
+  } else if (is.character(weights) && length(weights) == 1) {
+    # Weights as column name
+    weight_col <- weights
+    needed_cols <- c(formula_vars, weight_col)
+  } else if (inherits(weights, "formula")) {
+    # Weights as formula like ~cyl
+    weight_col <- all.vars(weights)
+    needed_cols <- c(formula_vars, weight_col)
+    # Store the extracted column name for later use
+    assign("weights_col", weight_col, envir = parent.frame())
+  } else if (is.numeric(weights)) {
+    # Weights as vector - store for later use
+    weight_col <- NULL
+    needed_cols <- formula_vars
+    assign("weights_vec", weights, envir = parent.frame())
+  } else {
+    stop("'weights' must be a column name, formula, or numeric vector", call. = FALSE)
+  }
+
+  # Extract needed columns
   data <- data[, .SD, .SDcols = needed_cols]
 
   lhs <- names(data)[1L]
-
   nobs_full <- nrow(data)
-
   data <- na.omit(data)
 
   # Convert columns of type "units" to numeric
@@ -393,13 +417,20 @@ model_response_ <- function(data, formula) {
 
 #' @title Check weights
 #' @description Checks if weights are valid
-#' @param x Regressor matrix
+#' @param y Dependent variable
+#' @param x Regressors matrix
 #' @param p Number of parameters
 #' @noRd
-check_linear_dependence_ <- function(x, p) {
-  if (qr(x)$rank < p) {
+check_linear_dependence_ <- function(y, x, p) {
+  # if (qr(x)$rank < p) {
+  #   stop("Linear dependent terms detected.", call. = FALSE)
+  # }
+
+  if (check_linear_dependence_svd_(y, x, p) == 1L) {
     stop("Linear dependent terms detected.", call. = FALSE)
   }
+
+  return(TRUE)
 }
 
 #' @title Check weights

R/felm.R

@@ -138,12 +138,22 @@ felm <- function(formula = NULL, data = NULL, weights = NULL, control = NULL) {
   model_response_(data, formula)
 
   # Check for linear dependence ----
-  check_linear_dependence_(cbind(y, x), p + 1L)
+  check_linear_dependence_(y, x, p + 1L)
 
   # Extract weights if required ----
   if (is.null(weights)) {
     wt <- rep(1.0, nt)
+  } else if (exists("weights_vec")) {
+    # Weights provided as vector
+    wt <- weights_vec
+    if (length(wt) != nrow(data)) {
+      stop("Length of weights vector must equal number of observations.", call. = FALSE)
+    }
+  } else if (exists("weights_col")) {
+    # Weights provided as formula - use the extracted column name
+    wt <- data[[weights_col]]
   } else {
+    # Weights provided as column name
     wt <- data[[weights]]
   }
 

R/fenegbin.R

@@ -155,7 +155,7 @@ fenegbin <- function(
   model_response_(data, formula)
 
   # Check for linear dependence in 'x' ----
-  check_linear_dependence_(cbind(y, x), p + 1L)
+  check_linear_dependence_(y, x, p + 1L)
 
   # Extract weights if required ----
   if (is.null(weights)) {