Statalist


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

Re: st: asclogit or clogit or mixlogit


From   Arne Risa Hole <arnehole@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: asclogit or clogit or mixlogit
Date   Thu, 4 Jun 2009 13:33:01 +0100

Hi Shehzad,

Your -clogit- model can be estimated using -asclogit- as follows:

bysort choiceset: gen alt = _n
asclogit choice frequency relief cost, case(choiceset) alt(alt) noconst

Look up the definition of the -alternatives()- option of -asclogit- in
the Stata manual. To capture preference heterogeneity in terms of the
observable characteristics of the decision-makers you need to interact
them with the alternative attributes. It makes no sense to include
them as covariates using the -casevars()- option since your
alternatives are unlabelled (what does it mean that female respondents
have a greater tendency to choose alternative A, for instance?)

-mixlogit- is a version of -asclogit- that allows the coefficients to
be random, whether you want this or not depends on the goal of your
research project.

Arne

2009/6/3 Shehzad Ali <drshehzad_ali@yahoo.com>:
> Hi listers,
>
> I am analysing a discrete choice experiment in which 100 individuals are asked about their treatment preference for pain relief. Each individual is offered 10 choice sets, each with two alternatives (A and B); each alternative has 3 attributes (frequency of treatment [frequency], duration of pain relief [relief] and cost [cost]). A typical choice set would look like this:
>
> Choice set # 1
>
> Alternative A:
> frequency of treatment (per day) = 3
> duration of pain relief (in hours) = 8
> cost ($) = 1
>
> Alternative B:
> frequency of treatment (per day) = 2
> duration of pain relief (in hours) = 12
> cost ($) = 1.5
>
> I also have individual level characteristics including age, sex, education and income. The data is in long table form; so I have 100 x 10 x 2 = 2,000 observations. The data is organised such that:
>
> id = remains the same for each individual, i.e. for each set of 20 observations
> choice = 0/1 dummy for choice made in each choice set
> choiceset = same for the two alternatives in each choice set, i.e. there are 1,000 choicesets numbered from 1 to 1,000
>
> I was using the conditional logit model below:
>
> clogit choice frequency relief cost, group(choiceset)
>
> Can anyone suggest whether using an -asclogit- or -mixlogit- be more appropriate in this situation, especially if I wish to take account of individual characteristics (I understand the other way would be to use interaction terms in -clogit-)? Also can you suggest the appropriate syntax for using -asclogit- and -mixlogit- in this situation (when I tried using: [asclogit choice frequency relief cost, casevars (age sex education income) case(id) alternatives(choiceset)], I get the error: variable choiceset has replicate levels for one or more cases).
>
> Any help would be greatly appreciated.
> Thank you,
> Shehzad
>
>
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index