When analyzing knowledge, one usually wants to match two regression fashions to find out which one suits finest to a chunk of information. Typically, one mannequin is a easier model of a extra complicated mannequin that features further parameters. Nevertheless, extra parameters don’t at all times 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 will 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 further parameters is significant or simply resulting from probability.
On this article I clarify the F-test for nested fashions after which I current a step-by-step algorithm, reveal its implementation utilizing pseudocode, and supply Matlab code that you could run instantly 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 features, on which I didn’t wish 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.
