Anova without intercept?

E

elad83

Guest
#1
Is it possible to perform Anova without an intercept?
In the software, there is an option to check the box that says “no intercept”. Once I do that, the test results become significant!
Is it okay to do that?
Thanks very much
 

Dason

Ambassador to the humans
#2
It's absolutely fine to do that. But if I were you I wouldn't since you don't seem to understand what it's really doing. The tests now being significant isn't really anything to get excited over because the test is testing something completely different.
 

Junes

New Member
#3
No. Only when you are absolutely sure your intercept is 0 in the population for theoretical reasons, and you know the relationship is strongly linear. See here. Which is practically never for most social science.

In general, it's a bad idea to look around for things to tweak to make your results significant.
 

Dason

Ambassador to the humans
#4
My understanding (although it could be incorrect since the software wasn't explicitly stated) was just that the 'no intercept' option for anova would fit a cell means model. Which is a perfectly valid way to parameterize the anova but like I said it doesn't sound like the OP is capable of interpreting the model as is.