0

im doing a model selection process in R

> drop1(test, test = "Chisq",)
Single term deletions

Model:
AFOTUs ~ Lev + SM + TSN + ANPP + (Lev:TSN)
        Df    AIC     LRT  Pr(>Chi)    
<none>     1455.4                      
SM       1 1478.6 25.2149 5.129e-07 ***
ANPP     1 1457.3  3.8631   0.04936 *  
Lev:TSN  1 1458.0  4.5681   0.03257 *

Here you can see there are 5 variables but when I use drop1()to select model I found this function can't distinguish between single and interactive variables.If a predictor is part of an interaction, it will not be dropped? So im wondering which function can deal with this situation.

yxh
  • 13
  • 2
  • It's not my area of expertise, but can you add a formula in the `scope=` of `drop1` to specify the output you desire? – thelatemail Jul 28 '21 at 02:52
  • 3
    This answer - https://stackoverflow.com/questions/41402592/why-is-drop1-ignoring-linear-terms-for-mixed-models/41402696#41402696 - looks to discuss this, and potentially why it might not be a good idea too. – thelatemail Jul 28 '21 at 03:14

0 Answers0