0

I have a data set from a survey of health problems and annoyance by traffic noise in people living in residential buildings, with "5-point scale answer" questions. I want to perform a multilevel model for an ordinal response (health problem or noise annoyance), and I tried with the gllamm command in Stata but it's not allowed the option link(ologit). Does anyone know another way to perform it? If not in Stata, in R or SPSS would be useful anyway.

StasK
  • 29,235
  • 2
  • 80
  • 165
Mauricio Fuentes
  • 3
  • 1
  • Would one of these help? - https://stat.ethz.ch/pipermail/r-sig-mixed-models/2010q1/003469.html. I won't call this an "answer" because I haven't done the work of going through them myself. – Peter Ellis Feb 02 '12 at 09:46
  • @Mauricio_Fuentes, welcome to the site. `ologit` is a valid link for `gllamm`. It must have been something else that was breaking down when you ran it. Can you show the code and the output? Alternatively, you can ask at statalist... although in all likelihood it will be me who will answer it there ;). Another Stata (user-written) command that may be useful is `reoprob`. – StasK Feb 18 '12 at 05:43
  • @StasK, I tried to do it with the example from "Multilevel and Longitudinal Modeling Using Stata", 1st Ed., by Rabe-Hesketh & Skrondal. Specifically, in Chapter 5, title 5.6.2: `gllamm impso weeksqrt treatment interact, i(id) link(ologit) adapt eform`, then Stata indicates `unknown link() ologit` `r(198);` – Mauricio Fuentes Feb 23 '12 at 02:20

1 Answers1

2

In all likelihood, you have an incomplete installation of gllamm. Type ssc install gllamm, replace to update your files.

StasK
  • 29,235
  • 2
  • 80
  • 165
  • I have no explanation for this, but I was completely fool. It´s rather obvious to try updating the installation. I did what you said, the command ran and I obtained the same results that those that appear in Rabe-Hesketh & Skrondal´s book. Thank you very much and see you soon! – Mauricio Fuentes Feb 29 '12 at 03:30
  • You might want to indicate that you accepted the answer, then, so that the discussion does not show up as open, and other people won't waste time looking at it. – StasK Feb 29 '12 at 15:30