Python library for conducting hypothesis and other group comparison tests.

The best way to install hypothetical is through pip.

pip install hypothetical

For those interested, the most recent development version of the library can also be installed by cloning or downloading the repo.

git clone [email protected]:aschleg/hypothetical.git
cd hypothetical
python setup.py install

• One-way Analysis of Variance (ANOVA)
• One-way Multivariate Analysis of Variance (MANOVA)
• Bartlett's Test for Homogenity of Variances
• Levene's Test for Homogenity of Variances
• Van Der Waerden's (normal scores) Test

• Chi-square test of independence
• Fisher's Exact Test
• McNemar's Test of paired nominal data
• Cochran's Q test
• D critical value (used in the Kolomogorov-Smirnov Goodness-of-Fit test).

• Chi-square statistic
• r (one-sample runs test and Wald-Wolfowitz runs test) statistic
• Mann-Whitney U-statistic
• Wilcoxon Rank Sum W-statistic

• Kurtosis
• Skewness
• Mean Absolute Deviation
• Pearson Correlation
• Spearman Correlation
• Covariance
  • Several algorithms for computing the covariance and covariance matrix of sample data are available
• Variance
  • Several algorithms are also available for computing variance.
• Simulation of Correlation Matrices
  • Multiple simulation algorithms are available for generating correlation matrices.

• Several algorithms for performing Factor Analysis are available, including principal components, principal factors, and iterated principal factors.

• Binomial Test
• t-test
  • paired, one and two sample testing

• Friedman's test for repeated measures
• Kruskal-Wallis (nonparametric equivalent of one-way ANOVA)
• Mann-Whitney (two sample nonparametric variant of t-test)
• Mood's Median test
• One-sample Runs Test
• Wald-Wolfowitz Two-Sample Runs Test
• Sign test of consistent differences between observation pairs
• Wald-Wolfowitz Two-Sample Runs test
• Wilcoxon Rank Sum Test (one sample nonparametric variant of paired and one-sample t-test)

• Chi-square one-sample goodness-of-fit
• Jarque-Bera test

• Tukey's Honestly Significant Difference (HSD)
• Games-Howell (nonparametric)

• Add noise to a correlation or other matrix
• Tie Correction for ranked variables
• Contingency table marginal sums
• Contingency table expected frequencies
• Runs and count of runs

The goal of the hypothetical library is to help bridge the gap in statistics and hypothesis testing capabilities of Python closer to that of R. Python has absolutely come a long way with several popular and amazing libraries that contain a myriad of statistics functions and methods, such as numpy, pandas, and scipy; however, it is my humble opinion that there is still more that can be done to make Python an even better language for data and statistics computation. Thus, it is my hope with the hypothetical library to build on top of the wonderful Python packages listed earlier and create an easy-to-use, feature complete, statistics library. At the end of the day, if the library helps a user learn more about statistics or get the information they need in an easy way, then I consider that all the success I need!

• Python 3.5+
• numpy>=1.13.0
• numpy_indexed>=0.3.5
• pandas>=0.22.0
• scipy>=1.1.0
• statsmodels>=0.9.0

MIT