Check to make sure there are no zeros in matched data. When we find them, bump them up based on the granularity. For now, figure 1/10 of if we had seen 1 in our calibration data, which is kind of unprincipled but better than crashing. Dan will review.
The model is in log space, and 0 is very bad.
Check to make sure there are no zeros in matched data. When we find them, bump them up based on the granularity. For now, figure 1/10 of if we had seen 1 in our calibration data, which is kind of unprincipled but better than crashing. Dan will review.