I'm not entirely sure what that minimum of 0 encodes -- is it saying that the output_type_ids for cdf model outputs must be at least 0 (if they are numeric)? If so, I think we should take that requirement out. Most disease forecasting systems would probably be set up with targets that are like "number of hospitalizations", in which case a minimum of 0 makes sense. But in principle, someone could have a different target like "change in number of hospitalizations", in which case negative values might be reasonable, corresponding to a decrease in hospitalizations.
Again, I'm not totally clear on what that entry is specifying, so please correct me if the above interpretation is wrong!
Looking at the schema for this line: https://github.com/hubverse-org/schemas/blob/47beca8c1c4e0e30ef6fcd079a9a74bfd8b24685/v4.0.0/tasks-schema.json#L941
I'm not entirely sure what that minimum of 0 encodes -- is it saying that the
output_type_ids
for cdf model outputs must be at least 0 (if they are numeric)? If so, I think we should take that requirement out. Most disease forecasting systems would probably be set up with targets that are like "number of hospitalizations", in which case a minimum of 0 makes sense. But in principle, someone could have a different target like "change in number of hospitalizations", in which case negative values might be reasonable, corresponding to a decrease in hospitalizations.Again, I'm not totally clear on what that entry is specifying, so please correct me if the above interpretation is wrong!