I am running a regression with a large number of factors and variables using the LFE command in the FELM package in R 3.4.3. When I include a large number of covariates and cluster the standard errors, I get the following error:
Error in Crowsum(xz, ia) : long vectors not supported yet: ../include/Rinlinedfuns.h:138
I think it is the implementation of the clustering. I have made testing, and my routine works just fine for a relatively large number of covariates (5-8k) without the clustering. The routine works without clustering with a small number of covariates (1k).
By looking at the source code, it seems that the construction of the clustered variance-covariance matrix breaks.
Any ideas?
Thanks in advance!