Thanks for your video. Really clear! Is there a guidance on how many clusters are needed for clustered standard errors to make sense? I tried running a regression with cov_type='clustered' and keep getting "covariance of constraints does not have full rank" warning, but this error does not exist when I just use the default nonrobust regression. Did some googling and literature is suggesting between 20-50 clusters. I only have ~ 10 clusters in my dataset. Is there an intuition on why there needs to have enough clusters for clustered standard errors to work? Or have I mis-understood the whole thing.
thank u so much！that really helps me a lot！ but i still have some confusion about cluster command in stata， for example， robust and vce cluster， should we use vce cluster preferentially？thx！
My pleasure! You might not want to include firm fixed effects when there is not enough variation in your data within each firm and therefore you need to test the variation across firms. You might use year fixed effects when there is variation over time that is not attributable to your other explanatory variables. A good example of where you would not want to use firm fixed effects, but would want to use year fixed effects is if you were testing the difference in stock performance at companies run by male CEO's and female CEO's and in your data the companies that were run by male CEO's were only ever run by male CEO's and the companies that were run by female CEO's were only ever run by female CEO's. In this case, you could not use firm fixed effects, but you could use year fixed effects to take out the effect of the overall market in given years. Let me know what other videos we can make that might help you!
I am attempting to identify the appropriate code for calculating clustered standard errors following the execution of a regression using the multinom() function in r. I attempted the following code but consistently encountered an error: Calculate the cluster-robust variance-covariance matrix vcov_clustered
thanks bro
Really good, thank you for your help!
Thanks for your video. Really clear! Is there a guidance on how many clusters are needed for clustered standard errors to make sense? I tried running a regression with cov_type='clustered' and keep getting "covariance of constraints does not have full rank" warning, but this error does not exist when I just use the default nonrobust regression. Did some googling and literature is suggesting between 20-50 clusters. I only have ~ 10 clusters in my dataset.
Is there an intuition on why there needs to have enough clusters for clustered standard errors to work? Or have I mis-understood the whole thing.
thank u so much！that really helps me a lot！ but i still have some confusion about cluster command in stata， for example， robust and vce cluster， should we use vce cluster preferentially？thx！
What would be a valid reason not to include firm fixed effects in a regression but year fixed effects? Thanks for your video by the way.
My pleasure! You might not want to include firm fixed effects when there is not enough variation in your data within each firm and therefore you need to test the variation across firms. You might use year fixed effects when there is variation over time that is not attributable to your other explanatory variables. A good example of where you would not want to use firm fixed effects, but would want to use year fixed effects is if you were testing the difference in stock performance at companies run by male CEO's and female CEO's and in your data the companies that were run by male CEO's were only ever run by male CEO's and the companies that were run by female CEO's were only ever run by female CEO's. In this case, you could not use firm fixed effects, but you could use year fixed effects to take out the effect of the overall market in given years. Let me know what other videos we can make that might help you!
I am attempting to identify the appropriate code for calculating clustered standard errors following the execution of a regression using the multinom() function in r.
I attempted the following code but consistently encountered an error:
Calculate the cluster-robust variance-covariance matrix
vcov_clustered