Conditional logistic regression in SPSS with CoxReg option

[Ange]

I have to do a conditional logistic regression with SPSS, and I´ve read in one post that this can be do thoughout the Cox Regression option.

The problem is that I´m not sure about the way I have to introduce the data.

I have a dependent variable (called "choice"), which can take the "value" A, B or C (3 options); 4 atributes (the independent variables, which can take 3 levels each), and the individuals (respondents).

Each respondent have to choose the preferred option among the three, and each option is characterized by having certain values of the different attributes, like in a "choice experiment".

I hope you could help me with the regression. Thanks in advance for your time and consideration!

[Bruce Weaver]

Hello Ange.  I have some questions.

1. Is the dependent variable purely nominal, or are the categories (A, B, C) ordered in some way?  

2. It's not clear why you need conditional logistic.  Do subjects respond more than once?  Or do you have matched pairs (or larger groups) of subjects?  

Thanks for clarifying.

If you do indeed have repeated measures (or matching), there are some options other than conditional logistic.  

- generalized estimating equations (GEE) available via GENLIN
- multilevel model with repeated measures at level 1, person/match-group at level 2 (via GENLINMIXED)

HTH.

Ange wrote

I have to do a conditional logistic regression with SPSS, and I´ve read in one post that this can be do thoughout the Cox Regression option.

The problem is that I´m not sure about the way I have to introduce the data.

I have a dependent variable (called "choice"), which can take the "value" A, B or C (3 options); 4 atributes (the independent variables, which can take 3 levels each), and the individuals (respondents).

Each respondent have to choose the preferred option among the three, and each option is characterized by having certain values of the different attributes, like in a "choice experiment".

I hope you could help me with the regression. Thanks in advance for your time and consideration!

[Ange]

There are different sets. Each set contains three options (A, B C). Each option has the same four attibutes but with different levels of them. For example:

OPtion A: atribute 1: has the lowest level; attribute 2: has the lowest level; attribute 3: has an intermediate level; and attrib. 4: has the highest level

OPTION B. A1: lowest level; A2: intermediate level; A3:intermediate level; A4: lowest level

and so on.

Each respondent values 3 sets (i.e selects for exaple : A, A, C).

I hope I´ve clear up your doubt!

Thanks

[Poes, Matthew Joseph-2]

I'm going to response to Bruces comment, as your direct questions to me require the same follow-up response as his does.  Answering his questions will help us help you.  I may have misunderstood what you were trying to do.  Do individual cases have changing attributes and/or make choices multiple times?  I'm going to assume I misunderstood your situation, and in fact there is a repeated design issue here, in which an individual makes multiple choices, and thus you need to account for the bias of the individual.  Given a subject variable (which you have) you would also need something to indicate each choice opportunity (which would replace the time variable).

I would never have thought to use Cox-regression in this way, but you certainly can use anything for a time variable which reflects the repeated measure.  In this way, I could certainly see that the repeat/time/choice variable could be used as the time variable in the cox regression.  I believe (but have certainly no direct experience with this) that it would then give you the information in which time point or choice point by attribute gives you a certain likelihood of choice.

I think that GEE could do the same thing, as Bruce mentions, and have a bit more experience with that.  In this case, again, you need to have the data stacked (a person is repeated for each choice they make) and you need to have a variable which indicates this.  Even if the sequential nature is unimportant, the number needs to sequence.  Genlinmixed can also work, as mentioned, and in fact, won't be so different from GEE.  If there are additional nesting issues to consider, beyond repeated choice in person, then use Genlinmixed.  If there are no other nesting structures, I'm not sure there would be any functional difference between the two.

Matthew J Poes
Research Data Specialist
Center for Prevention Research and Development
University of Illinois
510 Devonshire Dr.
Champaign, IL 61820
Phone:� 217-265-4576
email: [hidden email]



-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bruce Weaver
Sent: Friday, March 23, 2012 9:57 AM
To: [hidden email]
Subject: Re: Conditional logistic regression in SPSS with CoxReg option

Hello Ange.  I have some questions.

1. Is the dependent variable purely nominal, or are the categories (A, B, C) ordered in some way?

2. It's not clear why you need conditional logistic.  Do subjects respond more than once?  Or do you have matched pairs (or larger groups) of subjects?

Thanks for clarifying.

If you do indeed have repeated measures (or matching), there are some options other than conditional logistic.

- generalized estimating equations (GEE) available via GENLIN
- multilevel model with repeated measures at level 1, person/match-group at level 2 (via GENLINMIXED)

HTH.



Ange wrote

>
> I have to do a conditional logistic regression with SPSS, and I´ve
> read in one post that this can be do thoughout the Cox Regression option.
>
> The problem is that I´m not sure about the way I have to introduce the
> data.
>
> I have a dependent variable (called "choice"), which can take the "value"
> A, B or C (3 options); 4 atributes (the independent variables, which
> can take 3 levels each), and the individuals (respondents).
>
> Each respondent have to choose the preferred option among the three,
> and each option is characterized by having certain values of the
> different attributes, like in a "choice experiment".
>
> I hope you could help me with the regression. Thanks in advance for
> your time and consideration!
>

-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Conditional-logistic-regression-in-SPSS-with-CoxReg-option-tp5589618p5589702.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Anthony Babinec]

see

http://www.indiana.edu/~statmath/stat/all/cat/2c2.html

for sas and spss examples.

Tony Babinec
[hidden email]

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Ange
Sent: Friday, March 23, 2012 9:32 AM
To: [hidden email]
Subject: Conditional logistic regression in SPSS with CoxReg option

I have to do a conditional logistic regression with SPSS, and I´ve read in one post that this can be do thoughout the Cox Regression option.

The problem is that I´m not sure about the way I have to introduce the data.

I have a dependent variable (called "choice"), which can take the "value" A, B or C (3 options); 4 atributes (the independent variables, which can take 3 levels each), and the individuals (respondents).

Each respondent have to choose the preferred option among the three, and each option is characterized by having certain values of the different attributes, like in a "choice experiment".

I hope you could help me with the regression. Thanks in advance for your time and consideration!

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Conditional-logistic-regression-in-SPSS-with-CoxReg-option-tp5589618p5589618.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Ange]

Good expample! It represents quite good what I´m trying to do!

the problem (for me) remains: I have no idea of data insertion...

Because the example don´t explain it "step by step" (as I would need). Could you please help me?

Thanks!!

[Ange]

Following this example :

http://www.indiana.edu/~statmath/stat/all/cat/2c2.html 

suggested by Anthony Babinec, I´m going to try to simulate the data insertion, so you could correct me in case it´s necessary:

I have to create a variable to each attribute of the chocolat, and then, depending on the alternative, give them certain value (0 or 1).

I have to create a variable to represent the respondents called "id" for example.

I have to create a variable which is labelled (for example) "choice", which value is 1 if the respondent chooses the alternative, and 0 otherwise.

And finally I should create the variable "time", which is 1 if the respondent choses the alternative, an 2 otherwise:


Example of the resolution (imagine this table is introduced in SPSS):

id    kind  dark   nuts  soft   choice   t
1     1     0       0       0        1        1
1     2     1       1       0        0        2
1     3     0       1       0        0        2
1     4     0       0       1        0        2
1     5     0       1       1        0        2
1     6     1       1       1        0        2  
1     7     1       0       0        0        2
1     8     1       0       1        0        2


and so on ... (this represent the choice of one person, and the kind variable represents the kind of chocolate which is the "añternative" the respondent may choose)

To conduct the cos regression I should click: Analysis -> survival  -> cox regression; and then:

Time: I shoul give the "t" variable
Status: the variable choose (specifying that takes the value 1)
Covariables: dark, soft and nuts
Strata: I shoul indicate: "id" variable.

And that´s all??

Am I mistaken??? I have some doubts concerning the codification of the "t" variable...


Thanks!! Hope this "example" could clear up the missunderstandings...