How to bring the sample into a normal distribution?

Question

Python

Before using np.log() :


After using np.log() :


BUT

I want to test two groups with a T-Test, but first I decided to test for normality:

the result turned out to be very significant (i.e. the sample is not normally distributed)

How to get out of this to properly do a T-Test?

Answer 1

If your data is not normal, then there is little point in running tests based on the assumption that the data is normal.

Answer 2

"I now want to test two groups with a T-Test" - you can want anything. But this has nothing to do with data mining. If your data does not follow the normal distribution law, then using Student's t-test is extremely irrational. For such cases, there are other tests from the nonparametric group. A lot of them. You can start with the Wilcoxon-Mann-Whitney test, and/or the Kolmogorov-Smirnov test. Their standard implementations are in scipy.stats. However, as well as heaps of other methods that can be applied in such cases.