Sum of squares (of the erros) for each variable/modalities in a Proc GLM

[Mathis1]

Hello, 

I 'll try to be as clear as possible, because even in my head it's not very clear.

So, I performed a Proc GLM on a few categorical variables. What I'm looking to know is :

For the modality X1 (so when this modality is observed), what is the the sum of squares of the error. 

In other words, i'd like to have ∑ ( y – estimated(y)  ), for each time a modality is observed. 

On the output of the proc glm, i have "Type Error" for each modality but i don't know what it corresponds...

Any advice ?

Thank you !

[PaigeMiller · Accepted solution]

Sum of squares of the errors is defined only for the entire model. This is not a quantity that is computed (or defined) for an individual variable or level of a categorical variable.