Split val ==0 but se != 0

ihmeuw-msca / pyDisagg

pydisagg is a Python package for disaggregating estimated count observations across groups under generalized proportionality assumptions.

BSD 2-Clause "Simplified" License

1 stars 0 forks source link

Split val ==0 but se != 0 #52

Closed saalUW closed 4 months ago

saalUW commented 4 months ago

Doesn't split, matrix non inv. rate or log odds

zhengp0 commented 4 months ago

Hi @saalUW and @AHsu98, this is an interesting case. Currently we mark this as illegal https://github.com/ihmeuw-msca/pyDisagg/blob/main/src/pydisagg/ihme/splitter.py#L132 Along with #53, I am curious what makes this happen. For example, if the value to be split is prevalence and is 0, it means that there are 0 cases in the population they are considering, how do they measure se?