Ologit, Paralell Regression Assumption and weights

[Matteo Giuliani Pedraza]

Hi, it's the first time i write here, hope the question is not too emantary,

I'm trying to replicate a study from Gilkman and Semynov about ethnic segregation in Europe. They use data from the 2002 ESS and i'm using more recent data (2014) to see if something changed. In the original study they use an orderd logit having as dependent variable the perceived ethnic composition of the neighbourhood of residence (1-almost all residents are from minorities, 2- mixed, 3-Almost no residents are from minorities). I tried to use the same technique (first time i use it) in order to have comparable results, i used the ologit command but some problems came up.

-First, I tried to test for the Parallel Regression Assumption using the "brant" command, but it doesn't work with weights. I'm not sure if weights may alter the result of the test. Do you know other commands that could work with a weighted ordered model logit? Should I not worry about this?

-Second, when I use the "brant" command without weights, it tells me that the Parallel Regression Assumption has been violated. What alternative techniques would you advice?

Thankyou for your time and attention and thankyou in advance for your help!

[Richard Williams]

You can use the user-written gologit2 command for an alternative to brant when data are weighted. See

https://www3.nd.edu/~rwilliam/xsoc73994/Ologit01.pdf

especially p. 14.

gologit2 is also one means that can deal with violations. The support page is at

https://www3.nd.edu/~rwilliam/gologit2/index.html

Also, if your library has purchased access, you can read this article, which reviews several different ordinal regression models, including gologit2:

https://methods.sagepub.com/Foundati...ression-models

[Matteo Giuliani Pedraza]

Thankyou very much, this is very helpfull.

How ever, I still have a problem, it doesn't allow me to use the svy prefix, (wich, i understand, i should use since i'm using only data about migrants and only in certain countries). When I try to run the svy comand it returns the following error:

.         svy: gologit2 minorityneigh firstgen yearsmigration

gologit2 is not supported by svy with vce(linearized); see help svy estimation for a list of Stata estimation commands that are supported by svy

Looking online gsvy option, but it just gives me a different error:

.         gsvy: gologit2 minorityneigh firstgen yearsmigration

/ invalid name
an error occurred while attempting to compute scores

I don't know what i'm doing wrong.

Once again, thankyou very much for your help.

[Richard Williams]

My guess is you've got an outdated or corrupted version of gologit2. This is the most current:

. which gologit2
c:\ado\plus\g\gologit2.ado
*! version 3.2.5 17may2019 Richard Williams, [email protected]

If that isn't what you have, it is best to uninstall the old version first. Then type

ssc install gologit2, replace

If still having problems contact me directly.