https://github.com/fusion809/lrt-normal-nonconst-var
Likelihood-ratio test for normally distributed populations with non-constant variance. Essentially a LRT equivalent to Welch's ANOVA. Documented in LaTeX, implemented in Python.
https://github.com/fusion809/lrt-normal-nonconst-var
likelihood-ratio-test newtons-method statistics-in-python
Last synced: about 1 month ago
JSON representation
Likelihood-ratio test for normally distributed populations with non-constant variance. Essentially a LRT equivalent to Welch's ANOVA. Documented in LaTeX, implemented in Python.
- Host: GitHub
- URL: https://github.com/fusion809/lrt-normal-nonconst-var
- Owner: fusion809
- Created: 2021-09-19T14:02:30.000Z (over 4 years ago)
- Default Branch: master
- Last Pushed: 2021-09-30T10:38:11.000Z (over 4 years ago)
- Last Synced: 2025-01-27T10:11:30.352Z (over 1 year ago)
- Topics: likelihood-ratio-test, newtons-method, statistics-in-python
- Language: TeX
- Homepage:
- Size: 1.73 MB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Likelihood-ratio test for samples from normally distributed populations with non-constant variances
This repository contains a LaTeX document (including .tex file, .toc file and .pdf file output for it) showing the statistical working for a likelihood-ratio test that should be roughly equivalent to the Welch's ANOVA procedure.
The Python script [solver.py](/solver.py) implements the test. Currently it is set up to use ProjectData.csv, a file I, the author of this repository, have in my local copy of the repository that comes from the experimental project I did in [STA3300 at USQ in Semester 1, 2021](https://www.usq.edu.au/course/specification/2021/STA3300-S1-2021-ONC-TWMBA.html). This script uses Newton's method to derive the maximum likelihood estimators under the null hypothesis and uses an analytical formula derived in the LaTeX document in this repository for the unrestricted maximum likelihood.
[exOutput.txt](/exOutput.txt) contains the output from solver.py when run on my local copy of ProjectData.csv. The p-value obtained, 2.123e-09, is less than the p-value obtained by Welch's ANOVA (1.814e-06) and ordinary one-way ANOVA (1.314e-05), but higher than the p-value obtained by a Gamma generalized linear model (4.62e-13).