subsets of variables from unique combinations of larger set
Posted: Tue Sep 26, 2023 2:54 pm
With a total of n regressors, I would like to consider models (quantile regressions, actually) estimated separately for each unique combination of 1 to k variables, where k < n. For the k=2, it is trivial to use a pair of do loops for this. But in the more general case with k > 2, something more elegant is needed. Any suggestions for an efficient way to do this in RATS? (By way of explanation, this basically comes out of the forecasting-focused subset regression approach in Graham Elliott, Antonio Gargano, Allan Timmermann, Complete subset regressions, Journal of Econometrics, Volume 177, Issue 2, 2013, Pages 357-373, https://doi.org/10.1016/j.jeconom.2013.04.017.) Thanks much