Easy Hypothesis Test

A Python package for easy Hypothesis Tests

Authors

• @alod83

Installation

Install my-project with npm

  pip install easy-ht

Requirements

• scipy
• statsmodels
• jupyter-lab (optional)

Usage/Examples

For examples, check the folder examples, containing some Jupyter Notebooks to get started.

Documentation

The easy_ht package contains a basic class, called HypothesisTest.

HypothesisTest

A class used to calculate Hypothesis Tests, including both one sample and two sample tests.

Methods

• check_normality(self,alpha = 0.05) Check if samples are follow a normal distribution, using the Shapiro test.
• check_correlation(self, alpha = 0.05) Check if samples are correlated. It can be used only in two samples tests.
• check_randomness(self, alpha = 0.05, cutoff='mean') Check if the sample has been built in a random way.
• compare_means(self, value = None, alpha = 0.05, n = 50) In one sample test, compare the sample to an expected value. In two samples test, compare the mean of the two samples.
• compare_distributions(self, alpha = 0.05, cdf = None, args=(), freq = False) In one sample test, compare the sample to a distribution. In two samples tests, compare the distributions of the two samples.

init(self,x, y = None, verbose = False, alpha = 0.05):

Parameters

• x : array_like the (first) sample to be analysed

• y : array_like, optional the second sample to be analysed

• verbose : bool, optional, default = False enable debug messages

• alpha : float, optional, default = 0.05** the significance level

check_normality(self,alpha = 0.05)

Check if samples follow a normal distribution, according to the Shapiro test. In case of two samples, check if both the samples follow a normal distribution.

Parameters

• alpha : float, optional, default = 0.05 the significance level

Returns

• bool

True, if the sample of both the samples follow a normal distribution. False, otherwise.

check_correlation(self, alpha = 0.05)

Check if samples are correlated. If samples follow a normal distribution, the Pearson Correlation Coefficient is used, otherwise the Spearman Rank Correlation is used. This is a simple test, which does not return the statistics. Correlation is calculated only on the basis of p-value.

Parameters

• alpha : float, optional, default = 0.05 the significance level

Returns

• bool or None

True, if samples are correlated, False otherwise. None is returned in the case that the second sample has not been set.

check_randomness(self, alpha = 0.05, cutoff='mean'):

Check if the sample has been generated in a random way.

Parameters

• alpha : float, optional, default = 0.05 the significance level

• cutoff : {'mean', 'median'} or number, optional, default = 'mean' the cutoff to split the data into large and small values.

Returns

• bool

True, if the sample has been generated in a random way. False, otherwise.

compare_means(self, value = None, alpha = 0.05, n = 50)

Compare the sample mean to a theoretical value, or compare samples means. If samples follow a normal distribution, the t-test is used if the number of samples is less than n. The z-test, otherwise. If the samples are not normal, the Wilcoxon test is used.

Parameters

• value : float, optional the theoretical value to be compared, in case of one sample

• alpha : float, optional, default = 0.05 the significance level

• n : int, optional, default = 50 a number used to discriminate if a sample is small or big. if sample size <= n, t-test is used, otherwise z-test is used.

Returns

• bool True, if the sample means is similar to the theoretical value or the two samples means are similar. False, otherwise.

compare_distributions(self, alpha = 0.05, cdf = None, args=(), freq = False):

Compare the sample distribution to a given cdf (cumulative distribution function), if one sample is provided. The Kolmogorov-Smirnov Test is used. Compare the samples distribution, if two samples are provided. In this case, the Chi Square test is used.

Parameters

• alpha : float, optional, default = 0.05 the significance level

• cdf : str, array_like or callable if array_like, it is an array of observations of random variables, and the two-sample test is performed. If a callable, that callable is used to calculate the cdf. If a string, it should be the name of a distribution in scipy.stats, which will be used as the cdf function.

• args : tuple, sequence, optional distribution parameters, used cdf is string or callables.

• freq : bool, optional, default = False specify if the sample is an array of frequencies. This is used to discriminate if using the Chi Square Test or Kolmogorov-Smirnov Test.

Returns

• bool or None True, if the sample follows the specified distribution or the two samples follow the same distribution. False, otherwise. If error, return None.