Hi,
I have a question concerning the construction of a contrast to do a nested model comparison at the between subject level.
I ran a multiple regression with 4 predictors, including age, age^2, age^3, and gender for voxel wise prediction of grey matter concentration.
Now I would like to successively compare whether the extension by a quadratic or cubic age term is superior to the reduced model (with only the linear age effect), each time controlling for gender.
In other words, the test concerns whether the model
y = beta0 + beta1*age + beta2*gender
fits worse than the model
y = beta0 + beta1*age + beta2* age^2 + beta3*gender.
And in a further extension I would compare the latter model to the mode extended by age(cube).
My problem is how to generate the appropriate contrast for this. Any help would be appreciated.
Thank you,
Michael
|