Which analysis to use?

vl11

New Member
#1
Hi! I have a design with two categorical iv and one continuous iv with a continuous dv. I wanted to see if there was a main effect or interaction from the categorical ivs, so I used the syntax:

glm DV by IV1 IV2.

There were no significant interactions or main effects. I wanted to then add in the third and continuous iv, so I ran:

glm DV by IV1 IV2 with IV3/
design=IV1 IV2 IV3 IV1*IV3 IV2*IV3 IV1*IV2*IV3.

I'm confused if I did something wrong because the F and p values for IV1 and IV2 are different from the earlier output. Why is this and which one would I report?

Thanks so much!
 

EdGr

New Member
#2
Not sure what package you are using so I can't judge the syntax. But when you add one or more independent variables (to be adjusted for in the model) it is entirely possible that all the p-values without them would change. Suppose IV1 level 1 had high values of IV3, whereas IV1 level 2 had mostly low values of IV3. The model would adjust for that and could greatly reduce the significance of IV1 (assuming IV3 is associated with DV). This would be a correction for confounding by IV3, or it could be an overcorrection of some sort (say, IV1 changes values of IV3 in a causal way). In either case, p-values for IV1 could change.