Multiple regression with repeated measures (?)

[TomSnider]

I’d appreciate some opinions on how to analyze some data for a research project.  Each case in the analysis is a psychotherapist.  The dependent variable is a continuous measure of a certain client behavior, and we want to create a regression equation to predict the DV from three therapist characteristics while controlling for certain client characteristics (sex, age, etc.).  This seems to be a basic multiple regression situation, except that for each therapist there are three clients, and thus three DV values and three sets of client variables.

One simple approach would be to analyze only one (randomly selected) client from each therapist.  This would let us use the remaining clients as a hold-out sample for cross-validation.  But I wonder if there is a way to incorporate all the data into a single analysis.  

I’ll appreciate any thoughts on this!

[Valerie Villella]

Please note that I am off-site at at workshop on Tuesday November 20 to Wednesday November 21 and have limited access to my email. I will return to the office on Thursday November 22. If you require immediate assistance contact Nina Zivkovic at [hidden email].

Valerie Villella
Education Coordinator &
Policy and Program Analyst
OANHSS
905-851-8821 ext. 228

=====================
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

[Kylie]

Hi Tom,

If you think of each case being a client, rather than a therapist, then you appear to have a multilevel model with two levels (clients at level 1, therapists at level 2). Your dependent variable is measured at level 1 and you have a mix of level 1 and level 2 predictors. As you have a continuous outcome, see the MIXED procedure which can be used to fit multilevel models. The data would need to be structured with one row per client, with values for the therapist characteristics copied into each of the three client rows for each therapist.

Hope this helps.

Cheers,
Kylie.


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of TomSnider
Sent: Tuesday, 20 November 2012 9:37 AM
To: [hidden email]
Subject: Multiple regression with repeated measures (?)

I’d appreciate some opinions on how to analyze some data for a research
project.  Each case in the analysis is a psychotherapist.  The dependent
variable is a continuous measure of a certain client behavior, and we want
to create a regression equation to predict the DV from three therapist
characteristics while controlling for certain client characteristics (sex,
age, etc.).  This seems to be a basic multiple regression situation, except
that for each therapist there are three clients, and thus three DV values
and three sets of client variables.

One simple approach would be to analyze only one (randomly selected) client
from each therapist.  This would let us use the remaining clients as a
hold-out sample for cross-validation.  But I wonder if there is a way to
incorporate all the data into a single analysis.

I’ll appreciate any thoughts on this!




--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Multiple-regression-with-repeated-measures-tp5716330.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

[Bruce Weaver]

I concur with Kylie's advice.  If you need some good introductory material on these types of models, I recommend Jos Twisk's little orange book, "Applied Multilevel Analysis:  A Practical Guide for Medical Researchers".  Even if you're not a medical researcher, it's very accessible.  

http://books.google.ca/books/about/Applied_Multilevel_Analysis.html?id=N5nCQgAACAAJ&redir_esc=y

HTH.

Kylie Lange-3 wrote

Hi Tom,

If you think of each case being a client, rather than a therapist, then you appear to have a multilevel model with two levels (clients at level 1, therapists at level 2). Your dependent variable is measured at level 1 and you have a mix of level 1 and level 2 predictors. As you have a continuous outcome, see the MIXED procedure which can be used to fit multilevel models. The data would need to be structured with one row per client, with values for the therapist characteristics copied into each of the three client rows for each therapist.

Hope this helps.

Cheers,
Kylie.


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of TomSnider
Sent: Tuesday, 20 November 2012 9:37 AM
To: [hidden email]
Subject: Multiple regression with repeated measures (?)

I’d appreciate some opinions on how to analyze some data for a research
project.  Each case in the analysis is a psychotherapist.  The dependent
variable is a continuous measure of a certain client behavior, and we want
to create a regression equation to predict the DV from three therapist
characteristics while controlling for certain client characteristics (sex,
age, etc.).  This seems to be a basic multiple regression situation, except
that for each therapist there are three clients, and thus three DV values
and three sets of client variables.

One simple approach would be to analyze only one (randomly selected) client
from each therapist.  This would let us use the remaining clients as a
hold-out sample for cross-validation.  But I wonder if there is a way to
incorporate all the data into a single analysis.

I’ll appreciate any thoughts on this!




--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Multiple-regression-with-repeated-measures-tp5716330.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

[Ryan]

Do you realize that your name is nearly identical to a VERY well-known multilevel modeler?!:

 

http://www.stats.ox.ac.uk/~snijders/

 

With a name like yours, I have little doubt you'll be able to master this subject matter. :-)

 

All kidding aside, the other posters are obviously correct. I would suggest that you start with a random intercept model and treat all first-level and second-level predictors as fixed effects, and then add random slope terms as warranted.

 

Ryan

On Mon, Nov 19, 2012 at 6:06 PM, TomSnider <[hidden email]> wrote:

I’d appreciate some opinions on how to analyze some data for a research
project.  Each case in the analysis is a psychotherapist.  The dependent
variable is a continuous measure of a certain client behavior, and we want
to create a regression equation to predict the DV from three therapist
characteristics while controlling for certain client characteristics (sex,
age, etc.).  This seems to be a basic multiple regression situation, except
that for each therapist there are three clients, and thus three DV values
and three sets of client variables.

One simple approach would be to analyze only one (randomly selected) client
from each therapist.  This would let us use the remaining clients as a
hold-out sample for cross-validation.  But I wonder if there is a way to
incorporate all the data into a single analysis.

I’ll appreciate any thoughts on this!




--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Multiple-regression-with-repeated-measures-tp5716330.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

[statisticsdoc]

Tom,

I concur with the other posters that this appears to be a very suitable case for multi-level modeling, but with a few queries.

Do you have repeated measures for each client?  In which case, you would have a 3-level model:

Level 1: Time / Client / Therapist

Level 2:  Client / Therapist

Level 3: Therapist

How  many therapists do you have?

Are the patients randomized to different treatment conditions?  Are they randomized at the therapist level or within therapist?

SPSS Mixed could handle this analysis, although personally I like to cross check results with HLM software as well.

Best Regards,

Stephen Brand,  Ph.D.


www.StatisticsDoc.com


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of TomSnider
Sent: Monday, November 19, 2012 6:07 PM
To: [hidden email]
Subject: Multiple regression with repeated measures (?)

I’d appreciate some opinions on how to analyze some data for a research project.  Each case in the analysis is a psychotherapist.  The dependent variable is a continuous measure of a certain client behavior, and we want to create a regression equation to predict the DV from three therapist characteristics while controlling for certain client characteristics (sex, age, etc.).  This seems to be a basic multiple regression situation, except that for each therapist there are three clients, and thus three DV values and three sets of client variables.

One simple approach would be to analyze only one (randomly selected) client from each therapist.  This would let us use the remaining clients as a hold-out sample for cross-validation.  But I wonder if there is a way to incorporate all the data into a single analysis.

I’ll appreciate any thoughts on this!




--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Multiple-regression-with-repeated-measures-tp5716330.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