Develop hypothesis testing module(s)

[juanshishido]

Use case example: comparing nutrient uptake in plants.

Considerations:

• ensure randomized treatment
• number of plants being tested with each condition
• time series data (e.g., multiple "reads" throughout the plant's lifetime)—and how to analyze
• changing conditions across time—user initiated, for example
• more

[juanshishido]

"Traditional" hypothesis testing (z-test, t-test, and F-test, for example) assumes the data (or errors) are normally distributed.

If the normal assumption does not hold for the data and the sample size is small, the results of [these tests] are not reliable.

We may be able to transform the data so that this assumption holds. Otherwise, we can choose "nonparametric" tests—e.g., permutation tests. Nonparametric statistics make no assumptions about the probability distributions of the data.

Permutation tests often rival or even exceed the performance of parametric tests.

Sources

• Permutation Test & Monte Carlo Sampling
• CS 174, Lecture 11, John Canny