[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: St: Ordered Logit Question

From   "Jason Dean, Mr" <>
To   "" <>
Subject   RE: st: St: Ordered Logit Question
Date   Fri, 8 May 2009 17:00:56 -0400

Hmm, I'll have to take a look at it again. But when I run separate ologits and do predicted values they are much different then when I combine the sample and include a immigrant dummy (and of course interact all covariates). Maybe I made a typo in my code.

From: [] On Behalf Of Richard Williams []
Sent: Friday, May 08, 2009 5:41 PM
Subject: RE: st: St: Ordered Logit Question

At 03:20 PM 5/8/2009, Jason Dean, Mr wrote:
>Hi, you can include interation terms for the explanitory variables
>but not for the cut point (ancillary) paramaters. I find these to be
>much different between groups.

Not true, actually.  Simply including a dummy var for group
membership in ologit allows for index shift, e.g. if the coefficient
for group was 1, you could interpret that as saying all the cut
points were 1 point higher for one of the groups.

gologit2 (available from SSC) also allows for cut-point shift, i.e.
the cut-points can differ by non-constant amounts.  Maybe the first
cut point is one point higher, the 2nd is only .5 points higher,
etc.  Do something like

gologit2 y x1 x2 x3 group, npl(group)

See also pp. 14-20 of

As discussed earlier, there is the question of whether these group
coefficients reflect real effects or reflect differences in
measurement, i.e. groups use different cut points when reporting
their responses on the ordinal variable.  Of course, both might be
going on,  but if so I don't know how you would disentangle the two.

Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu

*   For searches and help try:
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index