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

Removed deep copy of the data in statistical test construction #218

Closed daryadedik closed 6 years ago

coveralls commented 6 years ago

Coverage Status

Coverage remained the same at 91.209% when pulling 825df813fc8fee1cf942af5b1d041b1dddd248ae on expan_data_bugfixes into c618b2d8c94e2a5e1f05e3dd0235d2ceae3f13c0 on master.

gbordyugov commented 6 years ago

what's the reason for the deep copy?

shansfolder commented 6 years ago

@gbordyugov since each StatisticalTest has a data object of its own, and we are passing the reference of same data frame around (in expan service), the initial idea of the deepcopy is to make each test own a clean copy of data.

shansfolder commented 6 years ago

@gbordyugov but then we run into memory issues because there are lots of tests for subgroups, deepcopy leads to memory error, thus we remove the copy in this PR.