Data which is not normal should be transformed so that it becomes normally distributed before applying parametric tests. Some of these transformations include:

- lognormal (or log base e or ln) (Y)
- 1 / Y
- square (Y)
- 1 / square (Y)
- square root (Y)
- 1 / square root (Y)

**Box-Cox transformation** is a convenient method that detects the deviation and applies the appropriate transformation to data. The package fifer provides a convenient function for this:

**code:**

> library(fifer)

> normalized_vect = boxcoxR(vect)

Where vect is a number series (vector).