New pred_only error function

[mnneely]

Sometimes users will want an output equation in the model for which there are no observations, e.g. output the amount in a compartment or the time above a threshold.

Please add a pred_only() dummy function in the $err list block which results in the following:

The corresponding output equation will generate predictions at idelta intervals without corresponding -99 observations in the data file. These can be returned with all other outputs in pred.csv.

[mhovd]

@Siel, don't we already support this out of the box?
I don't recall - do we exit or do we just warn if an output equation does not have a corresponding error model? We do have None as an option for it.

As for the observations, I don't think we are writing -99 anymore, if there are no observations then an empty field is used.
All output equations should have predictions made at idelta as well as observation timepoints already, so that one is free.

[Siel]

I was just having a conversation with professor Neely and so far I remember, we only generate simulated outputs for the output equations we have at least one observation.

The idea is to have a setting (in this case controlled by a dummy outeq) to let PMcore know whe want the predictions for that outeq to be generated.

[mhovd]

I think that the default should be to always generate predictions for all defined output equations, no?

[mnneely]

The idea of the (albeit dummy) pred_only() was to avoid an error if the number of output equations in the model is less than error equations, which is a check. It was also a way to be explicit that an output in the model was for predictions only.

[mhovd]

That should be fine - error polynomials are optional when generating predictions - they only affect the likelihood if provided.