# Hypothesis testing in Linear Discriminant Analysis

In order to check dependency between a categorical independent variable and a numerical dependent variable, one applies analysis of variance. If the dependency is vice versa, i.e., a categorical dependent variable and a numeric independent variable, which approach can I apply? For modeling and prediction purposes I can use Linear (Quadratic) Discriminant Analysis, but how can I check the hypothesis of independence?

• Was this helpful? Commented Mar 12, 2021 at 20:39

## 1 Answer

If I understand your question, then you want to know how to model a categorical dependent variable with a continuous independent variable.

In the case of a bivariate outcome, you use logistic regression. If you have more than two categories, you can consider multinomial regression. If your data is count data, then you would want to look at Poisson regression.

These are all types of General Linear Model (GLM), and will give you confidence intervals and pvalues.

Pretty much any statistical package should be able perform these for you. They are all very simple to do in R.

• I would add, you can also use LDA/MDA/ QDA or other algorithms such as SVM, decision tree, random forest, ANN, etc. but you need to understand the algorithm & its assumptions to be able to do Commented Mar 12, 2021 at 17:49