Dear All,
Thanks for your replies. I apologise for my lack of clarity in setting my
initial query. I will explain it in more detail now.
Basically, I have an ordered categorical model (ordered probit) where,
theoretically, people tend to overstate the values of the dependent variable
for higher values. That is, y can take for values 1, 2, 3 or 4 and my
assumption is that certain people in my sample (i.e., with some particular
values of covariates) tend to answer most often say, 3 or 4; so I am facing a
measurement error problem in the dependent variable, and by modelling the
cut-points as function of those covariates I would be able to take this into
account. However, as mentioned by Stat Kolenikov�s reply, I would face an
identification problem if the covariates I use to measure these errors in the
cut points are not different from the one in the main equation. The references
I found so far � not too many though- are:
Groot et al . Job satisfaction and preference drift. Economics Letters 63
(1999) 363�367.
Kristensen, N., Johansson, E. New evidence on cross-country differences in job
satisfaction
using anchoring vignettes. Labour Economics (2006).
Richard Williams suggested me to use gologit2 with constraints, but
unfortunately this is conceptually different to what I am trying to do: I want
to model the cut-points but not different betas for each category although this
adjustment might as well pin down the probabilities overestimation for p(y=3)
and p(y=4).
I repeat the questions of my previous post: a) Can I use glamm or goprobit to
estimate an ordered probit with cut points as function of covariates? b) Is it
any possibility that changing the ado file of oprobit could estimate it my
model or shall I program a new Maximum likelihood? In this case, does anybody
have any ML program for oprobit in which I can base to extent it? c) Does
anybody have any relevant literature about identification in this kind of
models?
Any further advice would be really welcome.
Thanks,
Marcos.
Quoting Richard Williams <[email protected]>:
> At 05:33 PM 4/10/2007, [email protected] wrote:
> >Assuming that sigma is one for easy notation, instead of estimating
> >the standard
> >probability for outcome j given by:
> >(1) Pr(y=j) = Phi(c_j-xb) - Phi(c_{j-1}-xb)
> >
> >I want to model c_j=f_j(z), where z is a subset of X:
> >(2) Pr(y=j) = Phi(f(z)-xb) - Phi(f_{j-1}(z)-xb)
> >
> >where f(z) could be modelled, for example, as: c_j=c_{j-1}+b*z, so the cut
> >points are funtions of z and have as contant the previous one. I almost sure
> I
> >cannot do that with the command oprobit, so I've been trying to use goprobit
> >and gllamm. However, I am not 100% confident that using either of these two
> >commands would do what I want to. My questions are:
> >a) Can I use glamm or goprobit to estimate equation (2)?
> >b) Is it any possibility that changing the ado file of oprobit could
> estimate
> >equation (2) or shall I program a new Maximum likelihood? In this case, does
> >anybody have any ML program for oprobit in which I can base to extent it?
> >c) Did anybody come across to same problem?
>
> I'm having a hard time visualizing exactly what you want, and
> why. Perhaps a hypothetical example, or a substantive explanation of
> when you would want to do this, would help.
>
> Both goprobit, and gologit2 with the -link(probit)- option, support
> the -constraints- option. So, if you can figure out how to specify
> what you want as a linear constraint, you might be able to do
> it. Here is a simple example:
>
> . use "http://www.indiana.edu/~jslsoc/stata/spex_data/ordwarm2.dta";
> (77 & 89 General Social Survey)
>
> . constraint 1 [#1]_cons = 1
>
> . constraint 2 [#2]_cons = .5
>
> . constraint 3 [#3]_cons=-.5
>
> . gologit2 warm yr89 male , link(p) c(1-3)
>
> Generalized Ordered Probit Estimates Number of obs =
> 2293
> Wald chi2(6) =
> 428.35
> Prob > chi2 =
> 0.0000
> Log likelihood = -2962.2844 Pseudo R2 =
> 0.0112
>
> ( 1) [1SD]_cons = 1
> ( 2) [2D]_cons = .5
> ( 3) [3A]_cons = -.5
> ------------------------------------------------------------------------------
> warm | Coef. Std. Err. z P>|z| [95% Conf.
> Interval]
> -------------+----------------------------------------------------------------
> 1SD |
> yr89 | .5751465 .0670129 8.58 0.000 .4438037
> .7064893
> male | -.1119678 .0501594 -2.23 0.026 -.2102784
> -.0136571
> _cons | 1 . . . .
> .
> -------------+----------------------------------------------------------------
> 2D |
> yr89 | .1486589 .0469871 3.16 0.002 .0565659
> .240752
> male | -.6175847 .0421769 -14.64 0.000 -.7002498
> -.5349195
> _cons | .5 . . . .
> .
> -------------+----------------------------------------------------------------
> 3A |
> yr89 | .0206849 .0514594 0.40 0.688 -.0801737
> .1215436
> male | -.754997 .0561626 -13.44 0.000 -.8650736
> -.6449203
> _cons | -.5 . . . .
> .
> ------------------------------------------------------------------------------
>
> Alternatively, I wonder if you want to selectively free some
> variables from the parallel lines constraint and then place
> constraints on their estimated coefficients. Again, a hypothetical
> example might help. Or, if there has been some published work that
> has done this, you could provide a citation.
>
>
> -------------------------------------------
> Richard Williams, Notre Dame Dept of Sociology
> OFFICE: (574)631-6668, (574)631-6463
> HOME: (574)289-5227
> EMAIL: [email protected]
> WWW: http://www.nd.edu/~rwilliam
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
