HYPOTHESIS TESTING

Introduction

Hypothesis testing is a statistical process closely related to model selection. However unlike model selection, hypothesis testing begins by adopting a default model - the null hypothesis (denoted H₀)- which is only abandoned when some alternative (denoted H₁) describes the data under consideration better.

This site aims to present different methods for performing common hypothesis tests, with focus upon the application of MML as outlined in detail in my thesis. For comparison, other approaches are also outlined.

Methods outlined are:

• Z-Test for a mean μ, when the standard deviation is known.
• χ²-Test for a standard deviation σ.
• T-Test for
a mean, when the standard deviation is unknown.
• F-Test for the comparison of two standard deviations σ₁ and σ₂.