Add Bootstrap for Hypothesis Testing

[pkaf]

Describe the workflow you want to enable

Add Bootstrap for hypothesis testing as shown in the latter part of https://en.wikipedia.org/wiki/Bootstrapping_(statistics)

Describe your proposed solution

Happy to code

Describe alternatives you've considered, if relevant

Aim is to extend hypothesis testing suite by providing a complementing method.

Additional context

Bootstrap is quite handy method in hypo testing allowing to provide probabilistic statement as to by what probability variant A is better or worse than B.

[rasbt]

Thanks, I would really appreciate additional features in MLxtend :). Btw. what's your view on using the bootstrap for hypothesis tests over permutation / randomization test? (btw. these are implemented here http://rasbt.github.io/mlxtend/user_guide/evaluate/permutation_test/). Usually, I hear people recommending bootstrap for confidence intervals only, and permutation / randomization tests for hypothesis testing. But of course, one can do a hypothesis test based on confidence intervals (i.e., if the confidence interval doesn't contain the null hypothesis, then the result is statistically significant at the given alpha level).

In either case, i'd be happy about that extra functionality. Maybe it can be implemented similar to http://rasbt.github.io/mlxtend/user_guide/evaluate/permutation_test/ for consistency

[pkaf]

Thanks.

Actually, I got the idea about adding bootstrap from your implementation of the permutation test. Will make it consistent to it.

Exactly as you described: if the confidence interval doesn't contain the null hypothesis, then the result is statistically significant at the given alpha level