Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: calculate probability for clogit
From
Steve Samuels <[email protected]>
To
[email protected]
Subject
Re: st: calculate probability for clogit
Date
Fri, 23 Apr 2010 10:08:21 -0400
I dropped car==3 to get two alternatives for the -asclogit- and
-clogit- example. As I said before, "choiceset" is not a proper case
variable; "pair" is, but will not, I speculate, give proper standard
errors. I suggest again that you consult somebody knowledgeable
about these analyses.
To answer your original question, I don't know how to get a proper
prediction for your design.
Good luck,
Steve
On Fri, Apr 23, 2010 at 9:05 AM, Patty_FARE_Guelph Apichaya
<[email protected]> wrote:
> Hi,
>
> Could you give me an example? How I can code it for 2 choice?
> I tried to code it but the program doesn't work.
>
> asclogit choice price brand labeling, case(choiceset) alternative( alt_choice)
> variable alt_choice has replicate levels for one or more cases; this is not allowed
>
> I look at your code and looked at how they code data in webuse choice, clear.
> drop if car==3, is it mean I have to generate alternative 3 in the data set? because I have only 2alternatives.
>
> However, when I look at the data set of -clogit- and -asclogit-. They are the same.
> clogit choice price brand labeling, group(choiceset)
>
> I agree that the case or group variable in -clogit- I will use pair (choiceset).
> For example, I have 32 choice set and 1 choice set has pair of alternatives (optionA and B)
> This is that way that I code in my data set :
>
> choiceset alternatives
> 1 1
> 1 2
> 2 1
> 2 2
> 3 1
> 3 2
> -
> -
> 32 1
> 32 2
> 1 1
>
> 1 2
>
> 2 1
>
> 2 2
>
> 3 1
>
> 3 2
>
> -
>
> -
>
> 32 1
>
> 32 2
>
>
> I will repeat the choice set again for next individual.
> Another point is should I continue choice set from 32 to 33 or should I repeat it from 1 to 32?
>
> Thank a lots
>
> Apichaya Lilavanichakul
> MSc. Candidate
> Department of Food Agricultural and Resource Economics
> Ontario Agricultural College
> University of Guelph
> Guelph, ON
>
>
>
>
>
> ----------------------------------------
>> Date: Fri, 23 Apr 2010 08:10:17 -0400
>> Subject: Re: st: calculate probability for clogit
>> From: [email protected]
>> To: [email protected]
>>
>> I should never post to Statalist in the middle of the night. I
>> apologize for the brevity of the last message.
>>
>> -asclogit- worked for me in Stata 11 with two outcomes using the
>> choice data set.
>>
>> ********************
>> webuse choice, clear
>> drop if car==3
>> asclogit choice , casevars(sex) case(id) alternatives(car)
>> tab car, gen(car)
>> gen car2_sex =car2*sex //add interaction of person-level variable
>> with alternative
>> clogit choice car dealer car2_sex , group(id) //sex enters only in
>> interaction
>> *********************
>> However I don't understand the pairwise design well-enough to be sure
>> of what you should do. One guess is that the case or group variable
>> should not be ID but "pair" and that to get proper standard errors you
>> should bootstrap the command with ID as a cluster variable for the
>> bootstrap. However, I think you need to consult somebody who knows
>> more about these analyses than I do. If you come back a correct
>> analysis in SAS, for example, we can probably help translate it into
>> Stata.
>>
>> Steve
>>
>> On Fri, Apr 23, 2010 at 12:54 AM, Steve Samuels wrote:
>>> Sorry, you are on your own. -asclogit- worked for me with two
>>> alternatives in Version 11 examples with the web "choice" data set.,
>>> but perhaps there's something special about your data, or you did not
>>> have the proper format (2 records per person and option). You can
>>> get person-level variables into -clogit- by adding their interactions
>>> with your alternative variable. However you cannot use the -clogit-
>>> data format for -asclogit-. As I said this is not my area, so read
>>> the manual entries.
>>>
>>> Good luck.
>>> Steve
>>>
>>> On Thu, Apr 22, 2010 at 9:34 PM, Patty_FARE_Guelph Apichaya
>>> wrote:
>>>> Hi,
>>>>
>>>> I try to use -asclogit- but it has to have more than 2 alternatives in 1 choice set, so I cannot use this command in my analysis.
>>>> For the choice design, I use SAS software to design the pair of alternative and combination of factors in order to balance and give the optimal sets. I use clogit to analyze my choices and I want to know how I calculate probability the people choose each alternative in choice set.
>>>>
>>>> if I want to calculate personal factors, should I use -xtlogit- or which command it is appropriate?
>>>>
>>>> Thanks
>>>>
>>>> Apichaya Lilavanichakul
>>>> MSc. Candidate
>>>> Department of Food Agricultural and Resource Economics
>>>> Ontario Agricultural College
>>>> University of Guelph
>>>> Guelph, ON
>>>>
>>>>
>>>>
>>>>> Date: Tue, 20 Apr 2010 16:12:44 -0400
>>>>> Subject: Re: st: calculate probability for clogit
>>>>> From: [email protected]
>>>>> To: [email protected]
>>>>>
>>>>> I know almost nothing about discrete choice experiments, so please
>>>>> read the following comments with that in mind.
>>>>>
>>>>> 1. -asclogit- might be a better choice than -clogit-, as it will more
>>>>> easily include person-level variables like gender, income, and sex.
>>>>>
>>>>> 2. Your setup does not identify the individual in any way. The
>>>>> examples for -asclogit- and -clogit- all use the individual as the
>>>>> case or group variable. As "set" has only four levels, perhaps you can
>>>>> include it as a person-level variable.
>>>>>
>>>>> 3. The examples in -asclogit- consider alternatives based on one
>>>>> attribute (factor). You have three factors and these generate 12 (=
>>>>> 3 x 2 x 2) possible options, with 66 possible pairwise comparisons.
>>>>> You selected 32 of these, divided into four sets. I'm curious--is
>>>>> there a reference for this experimental design? Is the choice of
>>>>> comparisons and construction of the sets optimal or balanced in some
>>>>> sense?
>>>>>
>>>>> 4. For predicted probabilities, use the -pr- option in -asclogit- .
>>>>>
>>>>> Steve
>>>>>
>>>>> On Tue, Apr 20, 2010 at 9:39 AM, Patty_FARE_Guelph Apichaya
>>>>> wrote:
>>>>>> Hi all,
>>>>>>
>>>>>> I use STATA to analyze discrete choihce experiment. My choice design has 32 choice sets with each 2 options (A and B). Each option include 3 factors: price (13,15,17), brand (AA,BB), labeling (animal,plant). The choice set will look like this.
>>>>>> Option A OptionB
>>>>>> price 13 15
>>>>>> brand AA BB
>>>>>> labeling animal plant
>>>>>>
>>>>>> From 32 choice sets is divide into 4 block with each 8 choice sets. Each individual (100 individuals) get 1 block (8 choice sets) and ask to choose between option A or B.
>>>>>>
>>>>>> choice = 1 chosen option, 0 not chosen option
>>>>>> set = same for two alternatives for each set i.e. there are 32 choice sets, it is repeated number from 1 to 32 and start it
>>>>>> again for the next 32 choice sets.
>>>>>>
>>>>>> This is the code; clogit choice price brand labeling, group(set)
>>>>>>
>>>>>> I also want to include individual characteristics; gender, income, age. how I can do it?
>>>>>>
>>>>>> How I can calculate the probability from people choosing choice?
>>>>>>
>>>>>> Any suggestion would be great.
>>>>>> Pat
>>>>>>
>>>>>>
>>>>>> _________________________________________________________________
>>>>>> Hotmail has tools for the New Busy. Search, chat and e-mail from your inbox.
>>>>>> http://www.windowslive.com/campaign/thenewbusy?ocid=PID28326::T:WLMTAGL:ON:WL:en-US:WM_HMP:042010_1
>>>>>> *
>>>>>> * 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/
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Steven Samuels
>>>>> [email protected]
>>>>> 18 Cantine's Island
>>>>> Saugerties NY 12477
>>>>> USA
>>>>> Voice: 845-246-0774
>>>>> Fax: 206-202-4783
>>>>>
>>>>> *
>>>>> * 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/
>>>>
>>>> _________________________________________________________________
>>>> The New Busy think 9 to 5 is a cute idea. Combine multiple calendars with Hotmail.
>>>> http://www.windowslive.com/campaign/thenewbusy?tile=multicalendar&ocid=PID28326::T:WLMTAGL:ON:WL:en-US:WM_HMP:042010_5
>>>> *
>>>> * 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/
>>>>
>>>
>>>
>>>
>>> --
>>> Steven Samuels
>>> [email protected]
>>> 18 Cantine's Island
>>> Saugerties NY 12477
>>> USA
>>> Voice: 845-246-0774
>>> Fax: 206-202-4783
>>>
>>
>>
>>
>> --
>> Steven Samuels
>> [email protected]
>> 18 Cantine's Island
>> Saugerties NY 12477
>> USA
>> Voice: 845-246-0774
>> Fax: 206-202-4783
>>
>> *
>> * 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/
>
> _________________________________________________________________
> The New Busy is not the old busy. Search, chat and e-mail from your inbox.
> http://www.windowslive.com/campaign/thenewbusy?ocid=PID28326::T:WLMTAGL:ON:WL:en-US:WM_HMP:042010_3
> *
> * 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/
>
--
Steven Samuels
[email protected]
18 Cantine's Island
Saugerties NY 12477
USA
Voice: 845-246-0774
Fax: 206-202-4783
*
* 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/