search

olshena / COVIDNearTerm

1 stars 0 forks source link

Adding debug mode #22

Open olshena opened 4 years ago

olshena commented 4 years ago

I would like to add a debug mode. This would be an argument to regressAR and simulateAR that defaults to FALSE. It would return as a separate object the phis and epsilons separately by training set and prediction set. It would also include the new lambdas that come into covariate prediction. It would be organized, probably in a matrix, so that it would be clears what phis and epsilons and lambdas went together.

It might be easier to do this first for simulateAR.

These are suggestions as to how to accomplish the goals, but I am open to better design. Please consider whether it should be a separate object.

kikapp commented 4 years ago

I've added a debug option to regressAR and simulateAR that outputs additional information when set to TRUE. Additional info is output in list form.

olshena commented 4 years ago

It seems like the debug includes only training set stats through the $buildAR_objects. I really like how this was done for the training set, but I would like it extended to the predictions. If it is there I can't see it.

kikapp commented 4 years ago

the debug mode for the prediction component of the algorithm is now implemented. it adds to run time significantly. now when debug = TRUE, predictAR returns a list with an item called predict_debug_object which contains a data frame that includes info from each iteration of each path in the algorithm

olshena commented 4 years ago

This looks good. It does take a long time to run. Can we output the training set debug parameters to the $buildAR_object by default? It is small so no reason to not do it.