Page 1 of 1

Binary outcome q-q plots for assumptions

Posted: Wed Mar 26, 2014 11:12 am
by KatHarvey1
Hi,
I have done tests of normality and heteroskedasticity for the level 1 data using q-q plots on MLwiN and they are coming up with two separate lines of residuals (next to each other with a break in the middle for normality, and on top of each other for heteroskedasticity). Has anyone had this happen before and if so is this normal? I wondered if it is showing the residuals for each of the two outcomes?
Any ideas would be much appreciated.
Thanks,
Kat

Re: Binary outcome q-q plots for assumptions

Posted: Wed Mar 26, 2014 5:21 pm
by joneskel
This quite usual and to be expected due to the binary nature of the response - see for single -level models

http://www.ats.ucla.edu/stat/stata/webb ... talog3.htm

In multilevel models, the q-q plots are more useful for the Normality assumption of the higher-level residuals

Re: Binary outcome q-q plots for assumptions

Posted: Thu Mar 27, 2014 4:20 pm
by KatHarvey1
That's really helpful. Thanks!

Re: Binary outcome q-q plots for assumptions

Posted: Tue Sep 21, 2021 10:18 pm
by nhighnote414
This is fairly common and to be expected given the binary character of the response - see for single-level models.
The q-q plots are more useful in multilevel models for the Normality assumption of the higher-level residuals.
That is really useful. Thanks!