Closed noxtoby closed 4 years ago
SuStaIn doesn't handle missing data at the moment, I have an idea about one way to go about it but it needs testing first
Yeah, I had a think about how to do it without biassing the model, but it's not as straightforward as it is for the EBM.
I close this as we have ongoing work on missing variable analysis in SuStaIn at the POND group, to be introduced later.
Any updates on handling missing data?
@d-morrison — see ZScoreSustainMissingData if using the z score model. If using mixture SustaIn, then you have to handle missing data yourself when calculating the event likelihoods that go into pySuStaIn.
I get divide by zero errors relating to model likelihood, which I tracked back to missing data causing problems with
max()
andmin()
, etc. Couldn't fix it withnumpy.nanmax()
, so we probably need to devise a robust method for handling missing data.