For people that have to compute and store a large variety of data and/or perform statistical inference. Check the tidypath and PhD-utils slides for an overview.
Don't spend time creating directories, deciding filenames, saving, loading, etc. Decorators savefig & savedata will do it for you with optimal compression. More info at the tidypath repository.
The module phdu.resample allows calls to the resample R package.
- Provides CI and permutation tests.
- CIs can account narrowness bias, skewness and other errors in CI estimation, as indicated in the article
- Alternatively, use
phdu.stats.bootstrapfor numba-accelerated computation (does not callresample).
Calculate the power for accepting H0 and estimate the needed sample size.
Function power_analysis in phdu.stats.bootstrap follows Efron-Tshibirani: An introduction to the bootstrap, p. 381-384.
Module phdu.stats.tests.permutation.
- Permutation tests for any statistic.
- Includes paired and block cases.
- tidypath and PhD-utils slides: instructions and use cases.
- Example notebook
-
For the R compatible installation first install R:
conda install -c conda-forge r r-essentials r-base -
Install with dependencies:
pip install phdu[dependencies]Where
dependenciescan bebase(recommended),all,r(needed forresampleto work),statsmodels,matplotliborplotly.