Skip to content

Fixes for KP stat calculation #612

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
lrberge merged 3 commits into from
Nov 25, 2025
Merged

Fixes for KP stat calculation #612

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
8fae41f
Fixes for KP stat
karldw Oct 19, 2025
2a5cc60
Clarify KP warning to list supported cases
karldw Oct 30, 2025
f7d91c6
In KP calc, return NA earlier
karldw Nov 24, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 16 additions & 7 deletions R/fitstats.R
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1619,12 +1619,23 @@ kp_stat = function(x){
# internal function => x must be a fixest object
#
# The code here is a translation of the ranktest.jl function from the Vcov.jl package
# from @matthieugomez (see https://github.com/matthieugomez/Vcov.jl)
#
# from @matthieugomez (see https://github.com/FixedEffects/Vcov.jl)


if(!isTRUE(x$iv) || !x$iv_stage == 2) return(NA)

n_endo = length(x$iv_first_stage)
n_inst = x$iv_n_inst
VCOV_TYPE = attr(x$cov.scaled, "type")
if (!identical(VCOV_TYPE, "IID") && n_inst != n_endo) {
warning(paste0(
"KP calculation is only implemented for the cases:\n",
" * vcov type is IID or\n",
" * number of endogenous variables == number of instruments"
))
return(NA)
}

# Necessary data

X_proj = as.matrix(resid(summary(x, stage = 1)))
Expand All @@ -1637,8 +1648,8 @@ kp_stat = function(x){
Z = model.matrix(x, type = "iv.inst")
Z_proj = proj_on_U(x, Z)

k = n_endo = ncol(X_proj)
l = n_inst = ncol(Z)
k = n_endo
l = n_inst

# We assume l >= k
q = min(k, l) - 1
Expand All @@ -1650,7 +1661,7 @@ kp_stat = function(x){
} else {
PI = coef(summary(x, stage = 1))
}
PI = PI[, colnames(PI) %in% x$iv_inst_names_xpd, drop = FALSE]
PI = as.matrix(PI[, colnames(PI) %in% x$iv_inst_names_xpd, drop = FALSE])

Fmat = chol(crossprod(Z_proj))
Gmat = chol(crossprod(X_proj))
Expand Down Expand Up @@ -1691,8 +1702,6 @@ kp_stat = function(x){
# There is need to compute the vcov specifically for this case
# We do it the same way as it was for x

VCOV_TYPE = attr(x$cov.scaled, "type")

if(identical(VCOV_TYPE, "IID")){
vlab = chol(tcrossprod(kronv) / nrow(X_proj))

Expand Down
7 changes: 6 additions & 1 deletion tests/fixest_tests.R
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2966,11 +2966,16 @@ names(base) = c("y", "x1", "x_endo_1", "x_inst_1", "fe")
set.seed(2)
base$x_inst_2 = 0.2 * base$y + 0.2 * base$x_endo_1 + rnorm(150, sd = 0.5)
base$x_endo_2 = 0.2 * base$y - 0.2 * base$x_inst_1 + rnorm(150, sd = 0.5)
base$cl_var = rep(1:50, 3) # using `fe` to cluster causes a vcov PSD warning

# Checking a basic estimation
est_iv = feols(y ~ x1 | x_endo_1 + x_endo_2 ~ x_inst_1 + x_inst_2, base)

fitstat(est_iv, ~ f + ivf + ivf2 + wald + ivwald + ivwald2 + wh + sargan + rmse + g + n + ll + sq.cor + r2)
fitstat(est_iv, ~ f + ivf + ivf2 + wald + ivwald + ivwald2 + wh + sargan + kpr + rmse + g + n + ll + sq.cor + r2)
test(c(length(est_iv$iv_first_stage), est_iv$iv_n_inst), c(2, 2)) # kpr relies on accessing these

est_iv_uneven = feols(y ~ x1 | x_endo_1 ~ x_inst_1 + x_inst_2, base, cluster = "cl_var")
fitstat(est_iv_uneven, ~ f + ivf + ivf2 + wald + ivwald + ivwald2 + wh + sargan + kpr + rmse + g + n + ll + sq.cor + r2)

est_fe = feols(y ~ x1 | fe, base)

Expand Down