Bug with model.matrix when no collinear fixed effects

[kylebutts]

Small bug found in model.matrix. If there are no collinear fixed effects, this fails. A first check to make sure that there are collinear variables can fix this. Something like: if(!"collin.var" %in% names(object))

fixest/R/MiscFuns.R

Lines 7887 to 7894 in 6c99dfe

	if(collin.rm){
	qui = which(colnames(linear.mat) %in% object$collin.var)
	if(length(qui) == ncol(linear.mat)){
	linear.mat = NULL
	} else if(length(qui) > 0){
	linear.mat = linear.mat[, -qui, drop = FALSE]
	}
	}

Reprex:

library(palmerpenguins)
library(fixest)

est <- feols(bill_length_mm ~ 0 | species, data = penguins)
#> NOTE: 2 observations removed because of NA values (LHS: 2).

model.matrix(est)
#> Error in if (length(qui) == ncol(linear.mat)) {: argument is of length zero

^{Created on 2021-11-09 by the reprex package (v2.0.0)}

[lrberge]

Thanks, fixed.