Closed martwo closed 1 year ago
The CDF should be constructed using float64 for best accuracy. Some datasets have their data field stored in float32. In that case the CDF could be float32, which might cause accuracy issues in the analysis, especially when there are many events in the MC dataset. https://github.com/icecube/skyllh/blob/dbafdf579f606914806a46e919789e1ccd417d4a/skyllh/core/random.py#L112-L113
The CDF should be constructed using float64 for best accuracy. Some datasets have their data field stored in float32. In that case the CDF could be float32, which might cause accuracy issues in the analysis, especially when there are many events in the MC dataset. https://github.com/icecube/skyllh/blob/dbafdf579f606914806a46e919789e1ccd417d4a/skyllh/core/random.py#L112-L113