I have two regression models.
lm(TEE ~ weight + gender)lm(TEE ~ BMR)
How do I compare these two models and check which one predicts TEE best?
0
$\begingroup$
$\endgroup$
1
$\begingroup$
$\endgroup$
Answering which model is "best" isn't very straightforward unfortunately, but here are a few metrics one can look at:
- Model with a higher R-square/adjusted r-square
- Better explanation of your test data
- Lower AIC value (Same training set used for both models, overfitting is unlikely here since you are using a maximum of two variables supported by enough number of observations.)
I'm also assuming the underlying relationship is linear in this case and that BLUE properties are verified for both models prior to this apples to apple comparison.
0
$\begingroup$
$\endgroup$
You could compare non-nested models by Davidson-Mackinnon test, it is implemented as jtest in lmtest package (http://www.inside-r.org/packages/cran/lmtest/docs/jtest)
