ordered logistic - diagnostics

[zuzannap]

Dear all,

What kind of model diagnostics do we perform in case of a multilevel ordered logistic regression model? By a continuous response variable, looking at qqplots and caterpillar plots we check whether the random effects at each level are normally distributed and homoscedastic and whether there are any outlying units. And what do we check in case of an ordinal response?

Bests,
Zuzanna

[zuzannap]

I posted this here as I (presumably incorrectly) attempted to add the term 'residuals(u)' to my runmlwin command for an ordinal response model and it does not work..

z

[ChrisCharlton]

Are you able to provide an example of your syntax, as the following works for me using the MLwiN sample data?:

Code: Select all

use "http://www.bristol.ac.uk/cmm/media/runmlwin/alevchem.dta", clear

generate gcseav = gcse_tot/gcse_no
egen gcseav_rank = rank(gcseav)
generate gcseav_uniform = (gcseav_rank - 0.5)/_N
generate gcseavnormal = invnorm(gcseav_uniform)
generate gcse2 = gcseavnormal^2
rename gender female
egen school = group(lea estab)

runmlwin a_point cons (gcseavnormal female gcse2, contrast(1/5)), ///
	level2(school: (cons, contrast(1/5)), residuals(u)) ///
	level1(pupil:) ///
	discrete(distribution(multinomial) link(ologit) denominator(cons) basecategory(6) pql2) nopause

[zuzannap]

Dear Chris,

It seems that I missed the second comma in expression "(cons, contrast(1/5)), residuals(u))". Now it works:) Thank you!

Bests,
Zuzanna