First, go to data view and filter the data to remove all the missing Xs. Then you can track down the singularity. I suspect you have a lot of missing data in one cross section and that's causing the problem. Still, you have an F stat in the last set of results, so that issue seems resolved.
Second, you are fine with an unbalanced panel (due to missing values).
Third, you can try multiple imputation. It will fill in the missing values (repeatedly) and possibly give you stronger results by increasing sample size.
Fourth, the coefficients are fairly large. Natural logs of level variables may be useful and could (or could not) tighten up your confidence intervals.
Fifth, I'm not sure about the data, but if you can get a prior year on RND which is lagged (for year 1997), you can increase your observation count. I've seen people lose lots of observations from lags even when data on the lagged variables are available though some of the other variables are not.
Second, you are fine with an unbalanced panel (due to missing values).
Third, you can try multiple imputation. It will fill in the missing values (repeatedly) and possibly give you stronger results by increasing sample size.
Fourth, the coefficients are fairly large. Natural logs of level variables may be useful and could (or could not) tighten up your confidence intervals.
Fifth, I'm not sure about the data, but if you can get a prior year on RND which is lagged (for year 1997), you can increase your observation count. I've seen people lose lots of observations from lags even when data on the lagged variables are available though some of the other variables are not.
Comment