The same recommenation for the linear regression models, n>> k+1, where k is the number of explanatory variables
the sample size should be enough to do the test of singnificance.
No, it is not enough to conduct parametric test (SFA), however you can go for TE through DEA.
You have a panel data.
How many explanatory variables?
You can fit the SFA model
José Francisco Moreira Pessanha Yes, I've Panel Data. Will my "n" be 9 or 90? What is the maximum number of explanatory variables I can use?
n is 90 (9 units x 10 years), maybe up to 2 or 3 explanatory variables
You can certainly run SFA as long as you have n>k+1. The question is more whether it is advisable to do so. My response to that is typically that yes you can, but the result of using small datasets is that you will have wide confidence intervals. Therefore if you run SFA, as long as you acknowledge the confidence intervals rather than just interpreting the TE term point estimate as a definite answer then it can still be a useful exercise.
I have in the past run SFA on different subsamples to get a sense for whether the estimate of TE is volatile (i.e. an inefficient estimator) and this can also be a good way of exploring the value of SFA with small datasets. As always with econometric analysis, these tools should only be used to help you understand the data better, not as a way of giving you de facto answers to questions without a detailed exploration of the data.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression