My question is essentially the same as in this thread, however the answer given wasn't especially exhaustive in that midas_r_ic_table must take the "table" argument, a table that can be generated via expand_weights_lags, another function that requires restrictions. Thus, even though midas_r_ic_table can output the model selection for an unrestricted model, to me it seems not possible to generate model selection where I have not already imposed a certain structure. Is there a way around this that I am not aware of?
If not, then my next question concerns model selection without midas_r_ic_table. Since midas_u is simply a wrapper for lm according to the documentation, I am wondering if there is a way to feed an unrestricted MIDAS model to existing model selection functions, or stepwise selection functions in R that are designed for linear models.
For context, I am fitting a RU-MIDAS model of the form :
mdl <- midas_r(mlsd(y,2,x)~mlsd(x, 1:3, x)+ mlsd(z, 1:3, x) +
mlsd(w, 1:3, x) +
+mlsd(y,3+0:3, x), data=list(dmx, dtx), start = NULL)
an example with reproducible code can be found here. It suffices to try running midas_r_ic_table on the fitted model without table arguments.
