How does one choose the best model based on ANOVA's result? I mean I have 3 model outputs 1st is linear+all interaction, 2nd is linear+pair wise interaction and 3rd is linear and I am asked to choose the best model out of those. What things should I look for making my decision? Is it residuals degree of freedom that I should pay attention to mostly? Please answer with explanation.
1
-
$\begingroup$ What was your goal in fitting the model? $\endgroup$– gung - Reinstate MonicaCommented May 25, 2015 at 18:03
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Some general approaches to start with: Command 'anova(mod1, mod2)' will show if these models are significantly different. Regression can be done and compared with R-squared, AIC or BIC. You can run stepwise regression and which variables / interactions are being included. RegBest function of FactoMineR package can be tried to get the best model. One can also divide the data set into training and testing part and compare different models. HTH.
