throw error if missing values due to pivoting long about sccomp HOT 3 CLOSED

for now missingness in the observed counts. this is indeed how we can identify outliers probabilistically.

missingness in the covariates would be awesome and it's not hard to implement. but this will be for the future

from sccomp.

Thanks @hansvancalster .

I think a parsimonious approach is to through error if values are missing. Actually, the model also allows for missing data, although this behaviour is not exported to the user.

Using tidyr::complete would fix the error.

from sccomp.

I agree with you. And indeed, tidyr::complete is what I used to circumvent the problem. Can you elaborate a bit on

the model also allows for missing data, although this behaviour is not exported to the user.

Is this about missingness in the response, in the predictor variables or both?

from sccomp.