I have successfully done type 2 and type 3 SS ANOVA on single models using several different packages.
I now need to perform ANCOVA, which means I need to compare two different models using the "anova" function - one with interaction and one without. Is there a way to perform "type III SS ANOVA" to compare the two different models?
Is there any need or sense to do such a thing? I have heard of performing type III ANOVA to test a model but not between TWO different models. My reason for wanting to perform type III is because I have unequal sample size in the factors (categorical variable).
Salvatore S. Mangiafico has been a tremendous help to me both is RG and through his very enlightening website on R and stats .
So if you are reading this Sir, do you have any advise?
Hi, Anand.
When you are constructing an anova table, you are essentially comparing different models. That is, the p-value for the interaction from the full model is the p-value from comparing the full model to the model without the interaction term.
The type III sum of squares is commensurate with comparing models with one term removed from the full model, without worrying about if that would leave higher order interactions without the lower terms in the model. I think. Or something like that.
Attached is an example showing that the p-value for the interaction from the full model is the p-value from comparing the full model to the model without the interaction term.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA