search

zalando / expan

Open-source Python library for statistical analysis of randomised control trials (A/B tests)
MIT License
334 stars 50 forks source link

Ensure that outlier detection works if there is NaN in the data #225

Closed aaron-mcdaid-zalando closed 6 years ago

aaron-mcdaid-zalando commented 6 years ago

In this PR, NaN will be replaced by the most negative floating point number prior to computing the percentile.

coveralls commented 6 years ago

Coverage Status

Coverage increased (+0.02%) to 92.461% when pulling e30f2609feebfd821cb51a0944da51d63f039853 on aaron-mcdaid-zalando:outlier.removal.with.NaNs into 08c600ba3601f839a633df3bdba0065fd549db6c on zalando:master.

gbordyugov commented 6 years ago

In this PR, NaN will be replaced by the most negative floating point number prior to computing the percentile.

Why exactly this solution and not smth else, for instance, dropping them altogether?