Hey,
I am currently conducting some empirical research where I run a regression using reghdfe. I use double clustered standard errors. My regression looks as follows:
I am interested in the coefficients on the x-variables. I have questions regarding the degrees of freedom for an analysis where I split my sample into different decades:
1. Am I right in assuming that the degrees of freedom used to check a single coefficient is 9 (since I have 10 years, minus 1)? Thus, the critical value for my t-stat is >3, although I have ~500k observations?
2. Is significance even really reliable in such a case with that little degrees of freedom? I am a little confused since I know a rule of thumb about the minimum number of observations, which is about 30 I think. But since I have 500k observations, I should be fine?
Thank you a lot in advance. I appreciate any help.
Best regards,
Jonathan
I am currently conducting some empirical research where I run a regression using reghdfe. I use double clustered standard errors. My regression looks as follows:
HTML Code:
reghdfe y x1 x2 x3 x4 x5, absorb(year) cluster(firm year)
I am interested in the coefficients on the x-variables. I have questions regarding the degrees of freedom for an analysis where I split my sample into different decades:
1. Am I right in assuming that the degrees of freedom used to check a single coefficient is 9 (since I have 10 years, minus 1)? Thus, the critical value for my t-stat is >3, although I have ~500k observations?
2. Is significance even really reliable in such a case with that little degrees of freedom? I am a little confused since I know a rule of thumb about the minimum number of observations, which is about 30 I think. But since I have 500k observations, I should be fine?
Thank you a lot in advance. I appreciate any help.
Best regards,
Jonathan
