I am trying to estimate parameters of a regression function, where:
- Independent Variables are continuous (and I might have up to 7 IVs)
- Dependent Variable is rank
The twist is that the sample is very small (5 observations at most); however, I do have the possibility of sampling multiple times, each time for a max of 5 observations.
What would be good practices here, regarding:
- Number of distinct samples?
- How to treat the ranks (1 thru 5) after sampling, in order to ensure comparability across the sub-samples?
- Knowing that the regression function is linear, what would be the best regression type to use?