Ranked ordered logistic regression (rologit) problems

[Soren Horndrup]

Hello Statlaist

I've some problems using the rologit command in Stata: I've collected data from a show in Italy, where each respondent has ranked the emotions (in dep var q1category_num) they tied to the show. Then I collected some background information about each respondent e.g. knowledge about the topic of the show (gknow), cultural consumption level (ccons) etc. I've converted the data in the long format, so that the 'norespondent' variable link each respondents rankings together. I want to estimate the influence of knowledge on how one ranks the emotions. I thought the rologit model would appropriate, but because there's no variation in all the independent variables within each respondent (it is a cross-sectional study) I cannot estimate the rologit. What do I do to estimate this function? I've converted the data into the long format to be able to have 1 dependent variable.

[Joseph Coveney]

It seems as if the only predictor you have is qlcategory_num. So, I think the command would be something like

rologit ranking_num i.qlcategory_num, group(norespondent)

According to the helpfile, if you want to inquire into respondent-invariant predictors, you form interaction terms of them with the varying predictors.

[Soren Horndrup]

Hi Joseph,

Thanks for your reply. So what I set out to do is to try to explain q1category_num with the predictor gknow (knwoledge) controlled for the rest of the variables. The norespondents variable is the grouping variable grouping together the alternatives ranked here. The ranking_num is just to see how order, but it is assumes by the rologit that it is ordered this way anyhow. What I'd like to do is to look for an between-group (respondents in this case) effect of the rankings - maybe there's another model more suited for this purpose? The problem to me, is to understand which model to use with a dependent variable that is nominal categories ranked-ordinally scaled?

[Joseph Coveney]

What I'd like to do is to look for an between-group (respondents in this case) effect o[n] the rankings

"According to the helpfile, if you want to inquire into respondent-invariant predictors, you form interaction terms of them with the varying predictors."

[Soren Horndrup]

I am still uncertain whether this model then makes much sense for my purpose. If I'd like to include a bunch of controls I'd end up with a lot a interactions, causing multicollinearity and making it hard for me to interpret the model...

[Joseph Coveney]

I can't help you about whether the general approach makes sense, but based on the title and contents of your original post you seemed to have settled on rologit. Have you tried it?

.ÿ
.ÿversionÿ15.1

.ÿ
.ÿclearÿ*

.ÿ
.ÿsetÿseedÿ`=strreverse("1495360")'

.ÿ
.ÿquietlyÿsetÿobsÿ250

.ÿgenerateÿintÿpidÿ=ÿ_n

.ÿ
.ÿforvaluesÿiÿ=ÿ1/5ÿ{
ÿÿ2.ÿÿÿÿÿÿÿÿÿgenerateÿbyteÿcontrol`i'ÿ=ÿruniform()ÿ>ÿ0.5
ÿÿ3.ÿ}

.ÿ
.ÿgenerateÿbyteÿknowledgeÿ=ÿruniform()ÿ>ÿ0.5

.ÿlabelÿdefineÿKnowledgesÿ0ÿGoodÿ1ÿExpert

.ÿlabelÿvaluesÿknowledgeÿKnowledges

.ÿ
.ÿquietlyÿexpandÿ4

.ÿbysortÿpid:ÿgenerateÿbyteÿitemÿ=ÿ_n

.ÿ
.ÿgenerateÿdoubleÿranduÿ=ÿruniform()

.ÿisidÿrandu

.ÿ
.ÿbysortÿpidÿ(randu):ÿgenerateÿbyteÿrankÿ=ÿ_n

.ÿ
.ÿrologitÿrankÿc.(control?)##i.itemÿc.knowledge##i.item,ÿgroup(pid)ÿnolog
control1ÿcontrol2ÿcontrol3ÿcontrol4ÿcontrol5ÿknowledgeÿomittedÿbecauseÿofÿnoÿwithin-pidÿvariance

Rank-orderedÿlogisticÿregressionÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿobsÿÿÿÿÿ=ÿÿÿÿÿÿ1,000
Groupÿvariable:ÿpidÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿgroupsÿÿ=ÿÿÿÿÿÿÿÿ250

NoÿtiesÿinÿdataÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿObsÿperÿgroup:
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿminÿ=ÿÿÿÿÿÿÿÿÿÿ4
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿavgÿ=ÿÿÿÿÿÿÿ4.00
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿmaxÿ=ÿÿÿÿÿÿÿÿÿÿ4

ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿLRÿchi2(21)ÿÿÿÿÿÿÿ=ÿÿÿÿÿÿ12.46
Logÿlikelihoodÿ=ÿ-788.2815ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿProbÿ>ÿchi2ÿÿÿÿÿÿÿ=ÿÿÿÿÿ0.9262

----------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿrankÿ|ÿÿÿÿÿÿCoef.ÿÿÿStd.ÿErr.ÿÿÿÿÿÿzÿÿÿÿP>|z|ÿÿÿÿÿ[95%ÿConf.ÿInterval]
-----------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿcontrol1ÿ|ÿÿÿÿÿÿÿÿÿÿ0ÿÿ(omitted)
ÿÿÿÿÿÿÿÿcontrol2ÿ|ÿÿÿÿÿÿÿÿÿÿ0ÿÿ(omitted)
ÿÿÿÿÿÿÿÿcontrol3ÿ|ÿÿÿÿÿÿÿÿÿÿ0ÿÿ(omitted)
ÿÿÿÿÿÿÿÿcontrol4ÿ|ÿÿÿÿÿÿÿÿÿÿ0ÿÿ(omitted)
ÿÿÿÿÿÿÿÿcontrol5ÿ|ÿÿÿÿÿÿÿÿÿÿ0ÿÿ(omitted)
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿÿÿÿÿÿitemÿ|
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ2ÿÿ|ÿÿ-.4728714ÿÿÿ.3018885ÿÿÿÿ-1.57ÿÿÿ0.117ÿÿÿÿ-1.064562ÿÿÿÿ.1188191
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ3ÿÿ|ÿÿ-.4217875ÿÿÿ.2988188ÿÿÿÿ-1.41ÿÿÿ0.158ÿÿÿÿ-1.007462ÿÿÿÿ.1638864
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ4ÿÿ|ÿÿ-.4479173ÿÿÿ.2998924ÿÿÿÿ-1.49ÿÿÿ0.135ÿÿÿÿ-1.035696ÿÿÿÿ.1398609
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿitem#c.control1ÿ|
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ2ÿÿ|ÿÿ-.0041206ÿÿÿ.2264133ÿÿÿÿ-0.02ÿÿÿ0.985ÿÿÿÿ-.4478826ÿÿÿÿ.4396414
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ3ÿÿ|ÿÿ-.2315427ÿÿÿ.2269485ÿÿÿÿ-1.02ÿÿÿ0.308ÿÿÿÿ-.6763535ÿÿÿÿ.2132681
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ4ÿÿ|ÿÿÿ.1404902ÿÿÿ.2259963ÿÿÿÿÿ0.62ÿÿÿ0.534ÿÿÿÿ-.3024543ÿÿÿÿ.5834347
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿitem#c.control2ÿ|
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ2ÿÿ|ÿÿÿ.1702194ÿÿÿ.2266677ÿÿÿÿÿ0.75ÿÿÿ0.453ÿÿÿÿÿ-.274041ÿÿÿÿ.6144799
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ3ÿÿ|ÿÿÿ.1007803ÿÿÿ.2283465ÿÿÿÿÿ0.44ÿÿÿ0.659ÿÿÿÿ-.3467706ÿÿÿÿ.5483312
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ4ÿÿ|ÿÿÿ.0441883ÿÿÿ.2260331ÿÿÿÿÿ0.20ÿÿÿ0.845ÿÿÿÿ-.3988285ÿÿÿÿÿ.487205
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿitem#c.control3ÿ|
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ2ÿÿ|ÿÿÿ.0480646ÿÿÿ.2292782ÿÿÿÿÿ0.21ÿÿÿ0.834ÿÿÿÿ-.4013123ÿÿÿÿ.4974416
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ3ÿÿ|ÿÿ-.0611784ÿÿÿ.2282258ÿÿÿÿ-0.27ÿÿÿ0.789ÿÿÿÿ-.5084928ÿÿÿÿÿ.386136
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ4ÿÿ|ÿÿ-.0554177ÿÿÿ.2278306ÿÿÿÿ-0.24ÿÿÿ0.808ÿÿÿÿ-.5019575ÿÿÿÿ.3911221
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿitem#c.control4ÿ|
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ2ÿÿ|ÿÿÿ.1148494ÿÿÿ.2275235ÿÿÿÿÿ0.50ÿÿÿ0.614ÿÿÿÿ-.3310885ÿÿÿÿ.5607874
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ3ÿÿ|ÿÿÿ.2306079ÿÿÿ.2271219ÿÿÿÿÿ1.02ÿÿÿ0.310ÿÿÿÿ-.2145429ÿÿÿÿ.6757587
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ4ÿÿ|ÿÿ-.0268236ÿÿÿ.2259349ÿÿÿÿ-0.12ÿÿÿ0.905ÿÿÿÿÿ-.469648ÿÿÿÿ.4160007
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿitem#c.control5ÿ|
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ2ÿÿ|ÿÿÿ.0919383ÿÿÿ.2276084ÿÿÿÿÿ0.40ÿÿÿ0.686ÿÿÿÿÿ-.354166ÿÿÿÿ.5380427
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ3ÿÿ|ÿÿÿ.1636866ÿÿÿ.2272714ÿÿÿÿÿ0.72ÿÿÿ0.471ÿÿÿÿ-.2817573ÿÿÿÿ.6091304
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ4ÿÿ|ÿÿÿ.0791843ÿÿÿ.2264784ÿÿÿÿÿ0.35ÿÿÿ0.727ÿÿÿÿ-.3647052ÿÿÿÿ.5230737
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿknowledgeÿ|ÿÿÿÿÿÿÿÿÿÿ0ÿÿ(omitted)
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
item#c.knowledgeÿ|
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ2ÿÿ|ÿÿÿ.3688725ÿÿÿ.2321429ÿÿÿÿÿ1.59ÿÿÿ0.112ÿÿÿÿ-.0861193ÿÿÿÿ.8238642
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ3ÿÿ|ÿÿÿ.4858897ÿÿÿ.2347734ÿÿÿÿÿ2.07ÿÿÿ0.038ÿÿÿÿÿ.0257424ÿÿÿÿ.9460371
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ4ÿÿ|ÿÿÿÿ.414582ÿÿÿÿ.231909ÿÿÿÿÿ1.79ÿÿÿ0.074ÿÿÿÿ-.0399512ÿÿÿÿ.8691152
----------------------------------------------------------------------------------

.ÿcontrastÿitem#c.knowledge,ÿnoestimcheck

Contrastsÿofÿmarginalÿlinearÿpredictions

Marginsÿÿÿÿÿÿ:ÿasbalanced

----------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿÿÿÿÿÿÿdfÿÿÿÿÿÿÿÿchi2ÿÿÿÿÿP>chi2
-----------------+----------------------------------
item#c.knowledgeÿ|ÿÿÿÿÿÿÿÿÿÿ3ÿÿÿÿÿÿÿÿ5.14ÿÿÿÿÿ0.1616
----------------------------------------------------

.ÿ
.ÿexit

endÿofÿdo-file


.