BookmarkSubscribeRSS Feed
0 Likes

Currently, PROC MIXED can implement just one type of heteroscedasticity consistent estimator of the variance, namely the Huber-White / empirical / sandwich / HC0 estimator; this is accomplished with the EMPIRICAL option on the PROC MIXED statement in conjunction with the SUBJECT= option on either the RANDOM or REPEATED statement:  https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_mixed_syntax01.htm#statug.mixed...

 

PROC REG, on the other hand, can implement 4 types of heteroscedasticity consistent estimators: HC0, HC1, HC2, and HC3 using the HCC option in the MODEL statement: https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_reg_syntax08.htm#statug.reg.mod...

 

Long and Ervin (2000) argue that HC3 is superior to HC0.

 

Additional methods such as HC4, HC4m, and HC5 have been proposed in the literature and implemented in the vcovHC function of R package sandwich

 

Please consider adding these additional heteroscedasticity consistent estimators to PROC MIXED.

 

Relevant references: 

Long J. S., Ervin L. H. (2000). “Using Heteroscedasticity Consistent Standard Errors in the Linear Regression Model.” The American Statistician, 54, 217–224.

Cribari-Neto F. (2004). “Asymptotic Inference under Heteroskedasticity of Unknown Form.” Comuputational Statistics & Data Analysis 45, 215–233.

Cribari-Neto F., Souza T.C., Vasconcellos, K.L.P. (2007). “Inference under Heteroskedasticity and Leveraged Data.” Communications in Statistics – Theory and Methods, 36, 1877–1888. Errata: 37, 3329–3330, 2008.

Cribari-Neto F., Da Silva W.B. (2011). “A New Heteroskedasticity-Consistent Covariance Matrix Estimator for the Linear Regression Model.” Advances in Statistical Analysis, 95(2), 129–146.