Hypothesis tests for uncertain values and collections

In addition to providing ensemble computation of basic statistic measures, this package also wraps various hypothesis tests from HypothesisTests.jl. This allows us to perform hypothesis testing on ensemble realisations of the data.

Terminology

Pooled statistics are computed by sampling all uncertain values comprising the dataset n times, pooling the values together and treating them as one variable, then computing the statistic.

Element-wise statistics are computed by sampling each uncertain value n times, keeping the data generated from each uncertain value separate. The statistics are the computed separately for each sample.

Implemented hypothesis tests

The following hypothesis tests are implemented for uncertain data types.

• One sample t-test.
• Equal variance t-test.
• Unequal variance t-test.
• Exact Kolmogorov-Smirnov test.
• Approximate two-sample Kolmogorov-Smirnov test.
• One-sample Anderson–Darling test.
• Jarque-Bera test.