When analyzing knowledge, one typically wants to check two regression fashions to find out which one matches finest to a bit of knowledge. Usually, one mannequin is a less complicated model of a extra complicated mannequin that features extra parameters. Nevertheless, extra parameters don’t all the time assure {that a} extra complicated mannequin is definitely higher, as they might merely overfit the info.
To find out whether or not the added complexity is statistically important, we are able to use what’s referred to as the F-test for nested fashions. This statistical method evaluates whether or not the discount within the Residual Sum of Squares (RSS) as a result of extra parameters is significant or simply as a result of likelihood.
On this article I clarify the F-test for nested fashions after which I current a step-by-step algorithm, show its implementation utilizing pseudocode, and supply Matlab code that you could run straight away or re-implement in your favourite system (right here I selected Matlab as a result of it gave me fast entry to statistics and becoming capabilities, on which I didn’t need to spend time). All through the article we’ll see examples of the F-test for nested fashions at work in a few settings together with some examples I constructed into the instance Matlab code.
