Using the fitlm function for model without intercept, how R-squared is being calculated?

The value of the R-squared when I define the linear model without intercept are mostly negative and somehow close to -3. When I apply the fitlm function for model without intercept, the R-squared values are between 0 and 1. Does anyone know which formula Matlab is using in this case?

[Shuqi Liu]

Also when using fitlm without intercept, the Rsquared comptued by fitlm is different using Matlab 2020a and prior vs 2020b. Does anyone knows why did it change?

Answer 1

[Rishabh Gupta]

Hi Farhad,

There are 2 types of R-squared values -

1. Ordinary R-Squared
2. Adjusted R-Squared

You can have a look at the formula in the below Link:

https://www.mathworks.com/help/stats/coefficient-of-determination-r-squared.html

Regards

[Pulkit Goyal]

This page doesn't explain the R-squared definitions when intercept is forced to zero.