Ordinal Multilevel Regression and Intra-class correlation

[Julia Ingenfeld]

Hello,

I am having some troubles with an ordinal multilevel regression. I have two levels and I would like to calculate the intra class correlation. After running my model, I used the estat icc command, but stata says "requested action not valid after most recent estimation command". I tried it with Stata 14 and 13.

I would calculate it by hand but in my output the individual-level variance is not reported. Did I understand it right that this is because it is fixed (I read something like this in a post about binary multilevel models)? Is there any way to claculate the ICC?

I'd be greatful for any help!

Best,
Julia

[Joseph Coveney]

I believe that the individual level variance is taken as 1 for ordered probit regression and as π² / 3 for ordered logistic regression. You can calculate ICC using the coefficient for the higher-level variance, which is returned from the fitted model as _b[var(_cons[higher-level varname]):_cons] for a conventional two-level model—illustrated below with an ordered probit regression.

.ÿversionÿ14.1

.ÿ
.ÿclearÿ*

.ÿsetÿmoreÿoff

.ÿsetÿseedÿ`=date("2015-12-09",ÿ"YMD")'

.ÿquietlyÿsetÿobsÿ250

.ÿ
.ÿgenerateÿintÿpidÿ=ÿ_n

.ÿ//ÿVarianceÿofÿ4ÿbelow,ÿsoÿICCÿshouldÿbeÿaboutÿ4ÿ/ÿ(1ÿ+ÿ4)ÿ=ÿ0.8
.ÿgenerateÿdoubleÿuÿ=ÿ2ÿ*ÿrnormal()

.ÿÿÿ
.ÿforvaluesÿtimeÿ=ÿ1/5ÿ{
ÿÿ2.ÿÿÿÿÿgenerateÿdoubleÿlatent`time'ÿ=ÿuÿ+ÿ(`time'ÿ-ÿ3)ÿ/ÿ5ÿ+ÿrnormal()
ÿÿ3.ÿ}

.ÿquietlyÿreshapeÿlongÿlatent,ÿi(pid)ÿj(time)

.ÿgenerateÿbyteÿresponseÿ=ÿ1

.ÿforvaluesÿcutÿ=ÿ1/4ÿ{
ÿÿ2.ÿÿÿÿÿquietlyÿreplaceÿresponseÿ=ÿresponseÿ+ÿ1ÿifÿnormal(latent)ÿ>=ÿ`cut'ÿ/ÿ5
ÿÿ3.ÿ}

.ÿ
.ÿmeoprobitÿresponseÿi.timeÿ||ÿpid:ÿ,ÿnolrtestÿnolog

Mixed-effectsÿoprobitÿregressionÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿobsÿÿÿÿÿ=ÿÿÿÿÿÿ1,250
Groupÿvariable:ÿÿÿÿÿÿÿÿÿÿÿÿÿpidÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿgroupsÿÿ=ÿÿÿÿÿÿÿÿ250

ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿObsÿperÿgroup:
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿminÿ=ÿÿÿÿÿÿÿÿÿÿ5
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿavgÿ=ÿÿÿÿÿÿÿÿ5.0
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿmaxÿ=ÿÿÿÿÿÿÿÿÿÿ5

Integrationÿmethod:ÿmvaghermiteÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿIntegrationÿpts.ÿÿ=ÿÿÿÿÿÿÿÿÿÿ7

ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿWaldÿchi2(4)ÿÿÿÿÿÿ=ÿÿÿÿÿÿ42.49
Logÿlikelihoodÿ=ÿ-1331.6879ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿProbÿ>ÿchi2ÿÿÿÿÿÿÿ=ÿÿÿÿÿ0.0000
------------------------------------------------------------------------------
ÿÿÿÿresponseÿ|ÿÿÿÿÿÿCoef.ÿÿÿStd.ÿErr.ÿÿÿÿÿÿzÿÿÿÿP>|z|ÿÿÿÿÿ[95%ÿConf.ÿInterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿÿtimeÿ|
ÿÿÿÿÿÿÿÿÿÿ2ÿÿ|ÿÿÿ.3668335ÿÿÿ.1230521ÿÿÿÿÿ2.98ÿÿÿ0.003ÿÿÿÿÿ.1256559ÿÿÿÿ.6080111
ÿÿÿÿÿÿÿÿÿÿ3ÿÿ|ÿÿÿÿ.267361ÿÿÿ.1252109ÿÿÿÿÿ2.14ÿÿÿ0.033ÿÿÿÿÿ.0219521ÿÿÿÿ.5127699
ÿÿÿÿÿÿÿÿÿÿ4ÿÿ|ÿÿÿ.7566463ÿÿÿ.1252324ÿÿÿÿÿ6.04ÿÿÿ0.000ÿÿÿÿÿ.5111953ÿÿÿÿ1.002097
ÿÿÿÿÿÿÿÿÿÿ5ÿÿ|ÿÿÿ.5771019ÿÿÿ.1251265ÿÿÿÿÿ4.61ÿÿÿ0.000ÿÿÿÿÿ.3318584ÿÿÿÿ.8223454
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿ/cut1ÿ|ÿÿ-.5359971ÿÿÿ.1623752ÿÿÿÿ-3.30ÿÿÿ0.001ÿÿÿÿ-.8542466ÿÿÿ-.2177476
ÿÿÿÿÿÿÿ/cut2ÿ|ÿÿÿ.0139792ÿÿÿ.1622471ÿÿÿÿÿ0.09ÿÿÿ0.931ÿÿÿÿ-.3040192ÿÿÿÿ.3319776
ÿÿÿÿÿÿÿ/cut3ÿ|ÿÿÿ.5194045ÿÿÿ.1630069ÿÿÿÿÿ3.19ÿÿÿ0.001ÿÿÿÿÿ.1999168ÿÿÿÿ.8388922
ÿÿÿÿÿÿÿ/cut4ÿ|ÿÿÿ1.065167ÿÿÿ.1647476ÿÿÿÿÿ6.47ÿÿÿ0.000ÿÿÿÿÿ.7422681ÿÿÿÿ1.388067
-------------+----------------------------------------------------------------
pidÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿvar(_cons)|ÿÿÿ4.330932ÿÿÿ.6659711ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ3.203991ÿÿÿÿ5.854251
------------------------------------------------------------------------------

.ÿ
.ÿdisplayÿinÿsmclÿasÿtextÿ"ICCÿ=ÿ"ÿ///
>ÿÿÿÿÿasÿresultÿ%04.2fÿ_b[var(_cons[pid]):_cons]ÿ/ÿ(1ÿ+ÿ_b[var(_cons[pid]):_cons])
ICCÿ=ÿ0.81

.ÿ
.ÿexit

endÿofÿdo-file


.

[Julia Ingenfeld]

Thank you!