How to get the sum of squares in GLM model summary ?

I want to calculate the % of variance explained by each variable in a GLM, for that I need the summ of square. Can we get it from the fitglm function output ?

Answer 1

[Ive J]

You can get SST, SSE and SSR:

mdl = fitglm(rand(100, 2), randi([0 1], 100, 1), 'dist', 'binomial', 'link', 'logit');

mdl.SSE

ans = 23.9711

mdl.SSR

ans = 1.0278

mdl.SST

ans = 24.9900

[Timbro]

thank you, but actally what I need is an ANOVA table as described here for lm : https://fr.mathworks.com/help/stats/linearmodel.anova.html

however, in my case the anova doesn't work :

modelspec = 'PSD_0_100Hz ~ Tide*Temperature*Salinity';
g = fitglm([Prof_BBG Temperature_BBG Salinite_BBG],Band_PSD_BBG_1_interp, modelspec, 'VarNames', {'Tide', 'Temperature', 'Salinity','PSD_0_100Hz'})
anova(g)

I get this message : "Check for incorrect argument data type or missing argument in call to function 'anova'."

[Ive J]

You cannot use anova on a GLM, but only an LM. The link you provided is the same. I assume your response is continuous, so you can use fitlm. Also, it's easier if you share your data.

[Timbro]

Indeed, the response is continuous, and using fitlm I can use anova

Thank you !