Recovering full variance-covariance matrix of estimated fixed effects

[justinekn]

Hi,

In a model with two dimensions of fixed effects (seller.f and buyer.f), I am interested in recovering Var(seller_i), Var(buyer_i), and Cov(seller_i, buyer_i) for all observations (i).

I have been able to recover Var(seller_i) and Var(buyer_i) thanks to the option se=TRUE which allows to recover the standard errors of the estimated fixed effects by bootstrap:

est3 <- felm(d_vat ~ period.f | seller.f + buyer.f, data=akm, keepX=TRUE)
beta3 <- getfe(est3, ef='zm2', se=TRUE, bN=100)

Can we recover the full var-covar matrix, in order to recover Cov(seller_i, buyer_i) for all observations ?