4x speed for cluster with purrr

pachadotdev · Feb 5, 2024 · 62b7a54 · 62b7a54
1 parent 76ed19f
commit 62b7a54
Show file tree

Hide file tree

Showing 13 changed files with 28 additions and 21 deletions.
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -16,6 +16,7 @@ Imports:
     Formula,
     magrittr,
     MASS,
+    purrr,
     rlang,
     stats
 Suggests: 

diff --git a/NAMESPACE b/NAMESPACE
@@ -45,6 +45,7 @@ importFrom(dplyr,ungroup)
 importFrom(dplyr,vars)
 importFrom(dplyr,where)
 importFrom(magrittr,"%>%")
+importFrom(purrr,map_dfr)
 importFrom(rlang,":=")
 importFrom(rlang,sym)
 importFrom(stats,as.formula)

diff --git a/R/capybara-package.R b/R/capybara-package.R
@@ -14,6 +14,7 @@
 #' @name capybara-package
 #' @importFrom dplyr all_of filter group_by mutate mutate_at select summarise
 #'  ungroup vars where
+#' @importFrom purrr map_dfr
 #' @importFrom rlang sym :=
 #' @importFrom Formula Formula
 #' @importFrom MASS negative.binomial theta.ml

diff --git a/R/generics_vcov.R b/R/generics_vcov.R
@@ -129,10 +129,9 @@ vcov.feglm <- function(
               cl <- cl.combn[, j]
               B.r <- B.r + crossprod(
                 as.matrix(
-                  do.call(
-                    rbind,
-                    lapply(split(G, G[, cl]), function(df) colSums(df[sp.vars]))
-                  )
+                  G %>%
+                    split(.[, cl]) %>%
+                    map_dfr(~ colSums(.x[sp.vars]))
                 )
               )
             }

diff --git a/docs/articles/intro.html b/docs/articles/intro.html
diff --git a/docs/pkgdown.yml b/docs/pkgdown.yml
@@ -3,5 +3,5 @@ pkgdown: 2.0.7
 pkgdown_sha: ~
 articles:
   intro: intro.html
-last_built: 2024-02-05T06:34Z
+last_built: 2024-02-05T19:03Z
 
diff --git a/docs/reference/apes.html b/docs/reference/apes.html
diff --git a/docs/reference/bias_corr.html b/docs/reference/bias_corr.html
diff --git a/docs/reference/feglm.html b/docs/reference/feglm.html
diff --git a/docs/reference/felm.html b/docs/reference/felm.html
diff --git a/docs/reference/fenegbin.html b/docs/reference/fenegbin.html
diff --git a/docs/reference/fepoisson.html b/docs/reference/fepoisson.html
diff --git a/vignettes/intro.Rmd b/vignettes/intro.Rmd
@@ -62,13 +62,12 @@ summary(fepoisson(
 ))
 ```
 
-The coefficients are almost idential to those in Table 3 from @Yotov2016 that
-were obtained with Stata.
+The coefficients are almost identical to those in Table 3 from
+@yotov2016advanced that were obtained with Stata. The difference is attributed
+to the different fitting algorithms used by the two software.
 
 To cluster the standard errors, we need to crete a pair variable and modify
-the model formula to add a cluster term. The result is similar and the
-difference is explained by the different algorithms used to estimate the
-variance-covariance matrix, which is based on @cameron2011robust.
+the model formula to add a cluster term.
 
 ```{r}
 dppml <- dppml %>%
@@ -83,4 +82,9 @@ summary(
 )
 ```
 
+The result is similar and the difference is explained by the different
+algorithms used to estimate the variance-covariance matrix. Capybara clustering
+algorithm is based on @cameron2011robust while Stata uses a sandwich-type 
+estimator.
+
 # References
-Original file line number
+Diff line change
@@ Expand Up / @@ -16,6 +16,7 @@ Imports: @@
         Formula,
         magrittr,
         MASS,
+        purrr,
         rlang,
         stats
     Suggests:
@@ Expand Down @@