Question about R2 for weighted linear regression in JMP

hongyu1

I used JMP to conduct a weighted linear regression analysis. However, RSquare value of the regression output by JMP (around 0.76) doesn't match my own calculation using the general formula of R2 = 1 - SS_residual/SS_total (around 0.78).

 

The regression parameters output by JMP match my own calculation, so I suppose the regression was conducted correctly. I was wondering if R2 output by JMP is incorrect, or JMP is using a different formula other than the general formula to calculate R2. 

 

https://en.wikipedia.org/wiki/Coefficient_of_determination

 

Attached is the input data I used for this analysis.

 

Screenshot 2024-08-09 at 2.28.07 PM.png

weighted_regression_data.csv

Victor_G

Hi @hongyu1,

 

You can check the formula used by JMP for the calculation of R² in this section : Summary of Fit

However, using this formula or the one you mentioned leads to identical results, as SS(total) = SS(model) + SS(error).

If I use your formula with the informations provided by JMP, I find the same R² value : 

1 - (SS(error) / SS(total)) = 1 - (1174,3486 / 4971,1282) = 1 - 0,23623382 = 0,763766 (as displayed by JMP on your screenshot).

 

Maybe the difference could be caused by a rounding in the calculation ?

 

Hope this answer will help you,