Feols: Incorrect Standard Errors for WLS with weights_type = "fweights" and vcov = "hetero" #367
Description
Example:
%load_ext autoreload
%autoreload 2
import pyfixest as pf
data = pf.get_data()
data_agg = data[["X1", "f1", "f2"]].groupby(["X1", "f1", "f2"]).size().to_frame("count").reset_index()
vcov = "hetero"
fit1 = pf.feols("X1 ~ f1", data = data, vcov = vcov)
fit2 = pf.feols("X1 ~ f1", data = data_agg, vcov = vcov, weights = "count", weights_type="fweights")
pf.etable([fit1, fit2], digits = 6)
"""
est1 est2
------------ ---------------------- ----------------------
depvar X1 X1
------------------------------------------------------------
Intercept 1.026086*** (0.049331) 1.026086*** (0.058791)
f1 0.001125 (0.002973) 0.001125 (0.003530)
------------------------------------------------------------
------------------------------------------------------------
R2 0.000147 -
S.E. type hetero hetero
Observations 998 998
------------------------------------------------------------
Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001
Format of coefficient cell:
Coefficient (Std. Error)
"""Note that this is not a problem for iid and CRV inference:
vcov = {"CRV1":"f2"}
fit1 = pf.feols("X1 ~ f1", data = data, vcov = vcov)
fit2 = pf.feols("X1 ~ f1", data = data_agg, vcov = vcov, weights = "count", weights_type="fweights")
pf.etable([fit1, fit2], digits = 6)
"""
est1 est2
------------ ---------------------- ----------------------
depvar X1 X1
------------------------------------------------------------
Intercept 1.026086*** (0.055146) 1.026086*** (0.055146)
f1 0.001125 (0.002904) 0.001125 (0.002904)
------------------------------------------------------------
------------------------------------------------------------
R2 0.000147 -
S.E. type by: f2 by: f2
Observations 998 998
------------------------------------------------------------
Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001
Format of coefficient cell:
Coefficient (Std. Error)
"""Metadata
Metadata
Assignees
Labels
Type
Projects
Status