10-22-2013 04:06 PM
I have an unbalanced panel dataset which consists of 434 observations of about 300 companies over 6 years. Among these companies, 197 of them only appear once in my dataset. The remaining companies appear at least twice in different time periods.
I know that PROC PANEL requires each cross sectional observation to appear at least in two different time periods and thus the 197 companies in my dataset which only appear once, can not be used in PROC PANEL. One of m colleagues suggested that instead of dropping those 197 observation, I run a random effects model with PROC MIXED. I am not sure if PROC MIXED has the same requirement for panel data regressions. Would you please let me know if I can retain those 197 companies and use PROC MIXED to run a random effects model with it? If yes, what is the rationale behind it? Why it is possible to use those observations in PROC MIXED and not PROC PANEL?
01-17-2017 04:09 PM