Re: st: What grouping should I use for clogit?

[Hadji Cortez Jalotjot <jalotjot@yahoo.com>]

Dear Klaus,

Sorry for this delayed response.

Anyway, to answer you questions.

Please pardon me if I am not able to explain the experimental design clearly.  I am quite new to this and still learning.  In fact I cannot fully understand the first question but I'll try to describe how I generated my experimental design for the choice experiment.  


For the choice sets, YES the respondents are presented with a 
multinomial choice.  For each choice set, there are 3 choices or 
options.  Then each respondents have 10 choice sets.

What I am trying to model is 
vehicle choice with the characteristics of the vehicle itself as the 
explanatory variables.  With this, the conditional logit model is most 
appropriate.
I already set up my data as required by the clogit command in Stata (shown in my previous email which I hope is clear enough).
My experimental design for this discrete choice experiment is quite simple.  An orthogonal starting design was created using R from a 3x2x5x3x2x3 possible combination of vehicle attribute levels.  30 rows of numbers were created.  Each row of the starting design represent one choice or option in a choice set.  The numbers in each row corresponds to certain attribute of a vehicle so each row is actually one vehicle.  To create the second choice or option, a foldover (adding 1 to the starting design modulo) was created.  So now I have 30 pairs of options.  I randomly divided the 30 pairs into 3 (10 pairs or choice sets).


In my data set, each choice set translates to 3 rows of observations (one row per each choice or option).  Since each respondent has 10 choice sets to answer, each respondent has a total of 30 rows of observations.

The clogit analyzes the data by groups.  If I only have one choice set per respondent, each respondent can be classified as one group.  But since I have 10 choice sets per respondent, I am a bit confused.  


If I set each respondent as the group, clogit will give a result (a good result that is) even though it says "note: multiple positive outcomes within groups encountered".  However, with the respondent as the "group" the predicted probabilities will be computed for the all the 30 options (10 x 3, sum of probabilities of the 30 choices/options is equal to 1) instead of predicting the probabilities for just 3 options in a choice set at a time.  I can remedy this by making each choice set as the group for clogit to analyze. 


My problem actually starts when I extend the model to include socio-economic characteristics of the respondent which is fixed for the same respondent.  If I do the remedy or solution I previously mentioned (grouping by choice set), will I be negating the fixed-effect since 10 choice sets were answered by the each respondent.

Hadji


________________________________
From: Klaus Pforr <kpforr@googlemail.com>
To: statalist@hsphsun2.harvard.edu 
Sent: Wednesday, January 25, 2012 6:53 PM
Subject: Re: st: What grouping should I use for clogit?

<>
Dear Hadji,

could you be a litte bit more specific with your experimtal design, at least with the literature? The point why I ask this is, that if you randomize the important parts, you dont need fixed effects models, and can use pooled or random effects models, which have can have more advantages, provided that you exogenize the important other variables with a good experimental design.

With regard to the statistical model, I'm still dont understand what the choice sets for each respondent are. Is this comparable to a multilevel model? To be more specific, do you have multinomial choice for your respondents over choice sets and choices?

I think, as you probably randomize the respondents, you can ignore individual fixed effects, and have to think about these choice sets, for which you probably could use fixed effects. This means you would use the choice set ids as the group var. But this rests on my guessing of your situation, so maybe I'm wrong.

best

Klaus

Am 25.01.2012 05:48, schrieb Hadji Cortez Jalotjot:
> Hi Klaus!
> 
> Thanks for the response.
> 
> 
> My experimental design is based on what is described by Bunch (1996). A starting design is generated.  I used R to generated an optimal
> design.  This starting design served as the first alternative in my choice
> experiment.  A foldover of the starting design is the second alternative.
>   Generic attributes were used.
> 
> 
> My
> dataset looks like this (this is for respondent 1):
>   respno - id of each respondent
> obsid - same number means it belong to
> the same choice set
> choice - dummy var YES or No
> the rest are the explanatory variables (attributes of a vehicle)
> 
> 
> respno obsid choice eng1_GAS eng2_DIE eng3_LPG eng4_HEV eng5_ELE pp1_700K pp2_1M pp3_more1M range1_more450 range2_450 range3_200 c100km1_400 c100km2_600 incentv1_WITH incentv2_NO emis1_same emis2_80 NONE
> 1 1 1 0 0 0 0 1 0 1 0 0 1 0 1 0 1 0 0 1 0
> 1 1 0 0 0 0 1 0 1 0 0 1 0 0 0 1 0 1 1 0 0
> 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
> 1 2 1 0 0 0 1 0 1 0 0 1 0 0 1 0 1 0 0 1 0
> 1 2 0 0 0 1 0 0 0 0 1 0 0 1 0 1 0 1 1 0 0
> 1 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
> 1 3 1 0 1 0 0 0 0 0 1 1 0 0 1 0 1 0 0 1 0
> 1 3 0 1 0 0 0 0 0 1 0 0 0 1 0 1 0 1 1 0 0
> 1 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
> 1 4 1 0 0 1 0 0 0 0 1 1 0 0 0 1 0 1 1 0 0
> 1 4 0 0 1 0 0 0 0 1 0 0 0 1 1 0 1 0 0 1 0
> 1 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
> 1 5 0 0 1 0 0 0 0 1 0 0 0 1 0 1 0 1 0 1 0
> 1 5 0 1 0 0 0 0 1 0 0 0 1 0 1 0 1 0 1 0 0
> 1 5 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
> 1 6 1 0 0 1 0 0 0 1 0 1 0 0 1 0 1 0 0 1 0
> 1 6 0 0 1 0 0 0 1 0 0 0 0 1 0 1 0 1 1 0 0
> 1 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
> 1 7 0 0 0 0 1 0 0 1 0 1 0 0 1 0 0 1 1 0 0
> 1 7 1 0 0 1 0 0 1 0 0 0 0 1 0 1 1 0 0 1 0
> 1 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
> 1 8 0 1 0 0 0 0 0 1 0 0 0 1 1 0 0 1 1 0 0
> 1 8 1 0 0 0 0 1 1 0 0 0 1 0 0 1 1 0 0 1 0
> 1 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
> 1 9 0 1 0 0 0 0 0 0 1 0 1 0 1 0 1 0 1 0 0
> 1 9 1 0 0 0 0 1 0 1 0 1 0 0 0 1 0 1 0 1 0
> 1 9 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
> 1 10 1 0 0 1 0 0 0 1 0 0 1 0 0 1 1 0 0 1 0
> 1 10 0 0 1 0 0 0 1 0 0 1 0 0 1 0 0 1 1 0 0
> 1 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
> 
> 
> My dependent variable is choice.
>   I understand clogit analyzes the data by
> groups.
>   So which should I assign as group? Each
> individual is the group (respno) OR each choice experiment (obsid)?
>   Thanks!
>   Hadji
> 
> 
> 
> 
> ----- Original Message -----
> From: Klaus Pforr<kpforr@googlemail.com>
> To: statalist@hsphsun2.harvard.edu
> Cc:
> Sent: Tuesday, January 24, 2012 8:11 PM
> Subject: Re: st: What grouping should I use for clogit?
> 
> Dear Hadji,
> 
> could tell a little bit more about your data, and the experimental
> design? A data example would be helpful.
> 
> best
> 
> Klaus
> 
> Am 23.01.2012 07:46, schrieb Hadji Cortez Jalotjot:
>> I am estimating a choice model using the clogit command.  Data is from a choice experiment.
>> 
>> Each respondent was shown 10 choice sets.  Each choice set has 3 choices.
>> 
>> So basically, each respondent corresponds to 30 rows of observations.
>> 
>> My question is, should the respondents (respno) be my group for the clogit?  or should I treat each choice as the "group"?
>> 
>> Any suggestions or explanations are appreciated.
>> 
>> Hadji
>> 
>> 
>> *
>> *   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/
> 


-- __________________________________

Klaus Pforr
MZES AB - A
Universität Mannheim
D - 68131 Mannheim
Tel:  +49-621-181 2797
fax:  +49-621-181 2803
URL:  http://www.mzes.uni-mannheim.de

Besucheranschrift: A5, Raum A309
__________________________________

*
*   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/