Doubly-multivariate Repeated Measures Design

Doubly-multivariate Repeated Measures Design

Post by Leheman » Tue, 31 Oct 2000 06:46:51



hi, here is my problem:

my data contains 2 class variables -- Denture Treatment (2 levels, A &
B) and
Insulin Treatment (2 levels, 0 for No & 1 for Yes). My response
variables are
measured repeatedly for both before Denture Treatment & after Denture
Treatment.

In SASonlinedoc, an example similar with mine is like this:
( \onlinedoc\sasdoc\sashtml\stat\chap30\sect58.htm
SAS Release 8.1)
=========================================
data Trial;
      input Treatment $ Repetition PreY1 PostY1 FollowY1
                                   PreY2 PostY2 FollowY2;
      datalines;
   A        1  3  13  9  0  0  9
   A        2  0  14 10  6  6  3
   A        3  4   6 17  8  2  6
   A        4  7   7 13  7  6  4
   A        5  3  12 11  6 12  6
   A        6 10  14  8 13  3  8
   B        1  9  11 17  8 11 27
   B        2  4  16 13  9  3 26
   B        3  8  10  9 12  0 18
   B        4  5   9 13  3  0 14
   B        5  0  15 11  3  0 25
   B        6  4  11 14  4  2  9
   Control  1 10  12 15  4  3  7
   Control  2  2   8 12  8  7 20
   Control  3  4   9 10  2  0 10
   Control  4 10   8  8  5  8 14
   Control  5 11  11 11  1  0 11
   Control  6  1  5  15  8  9 10
   ;

   proc glm data=Trial;
      class Treatment;
      model PreY1 PostY1 FollowY1
            PreY2 PostY2 FollowY2 = Treatment / nouni;
      repeated Response 2 identity, Time 3;
   run;
=========================================

For my problem, I just need to add another Class variable in the
class statement. BUT, it seems in this analysis, the Response
variables need to be Continuous data, while my Reponse are
Ordinal data (from multiple-choice questions of sampling paper)

So, is there any other way to do this analysis suitable for my
data?

 
 
 

Doubly-multivariate Repeated Measures Design

Post by Michael Babya » Tue, 31 Oct 2000 08:01:57


: For my problem, I just need to add another Class variable in the
: class statement. BUT, it seems in this analysis, the Response
: variables need to be Continuous data, while my Reponse are
: Ordinal data (from multiple-choice questions of sampling paper)

: So, is there any other way to do this analysis suitable for my
: data?

It depends on the nature of the multiple choice questions.  If a
Likert-type item is scaled poperly (or if you understand the scaling
properties very well), there's really no harm in analyzing it as if it
were a continuous variable.  If you are summing several Likert-type items,
which would be more desirable than analyzing separate items, there is even
less of a problem; again, this assumes that the items were constructed
carefully.

Mike Babyak

--
_________________________________________________________________
  Michael A. Babyak, PhD                (919) 684-8843 (Voice)  
  Box 3119                              (919) 684-8629 (Fax)            
  Department of Psychiatry                                      

  Durham, NC 27710                                              
_________________________________________________________________
  In Flumine Stercoris Noli Undas Facere                        
_________________________________________________________________

 
 
 

Doubly-multivariate Repeated Measures Design

Post by Leheman » Tue, 31 Oct 2000 12:00:14


For my problem, the each of the Responses is a sum score of several (3-5)
Likert-type scale questions after using Cluster Analysis. Each question
has score 1, 2, 3 ,4. So my responses have ranges about 10. It seems
I can treat them as continuous data.

Thanks a lot!



> : For my problem, I just need to add another Class variable in the
> : class statement. BUT, it seems in this analysis, the Response
> : variables need to be Continuous data, while my Reponse are
> : Ordinal data (from multiple-choice questions of sampling paper)

> : So, is there any other way to do this analysis suitable for my
> : data?

> It depends on the nature of the multiple choice questions.  If a
> Likert-type item is scaled poperly (or if you understand the scaling
> properties very well), there's really no harm in analyzing it as if it
> were a continuous variable.  If you are summing several Likert-type items,
> which would be more desirable than analyzing separate items, there is even
> less of a problem; again, this assumes that the items were constructed
> carefully.

> Mike Babyak

 
 
 

Doubly-multivariate Repeated Measures Design

Post by John Jon » Tue, 31 Oct 2000 22:12:48


Proc CatMod?


Quote:>hi, here is my problem:

>my data contains 2 class variables -- Denture Treatment (2 levels, A &
>B) and
>Insulin Treatment (2 levels, 0 for No & 1 for Yes). My response
>variables are
>measured repeatedly for both before Denture Treatment & after Denture
>Treatment.

>In SASonlinedoc, an example similar with mine is like this:
>( \onlinedoc\sasdoc\sashtml\stat\chap30\sect58.htm
>SAS Release 8.1)
>=========================================
>data Trial;
>       input Treatment $ Repetition PreY1 PostY1 FollowY1
>                                    PreY2 PostY2 FollowY2;
>       datalines;
>    A        1  3  13  9  0  0  9
>    A        2  0  14 10  6  6  3
>    A        3  4   6 17  8  2  6
>    A        4  7   7 13  7  6  4
>    A        5  3  12 11  6 12  6
>    A        6 10  14  8 13  3  8
>    B        1  9  11 17  8 11 27
>    B        2  4  16 13  9  3 26
>    B        3  8  10  9 12  0 18
>    B        4  5   9 13  3  0 14
>    B        5  0  15 11  3  0 25
>    B        6  4  11 14  4  2  9
>    Control  1 10  12 15  4  3  7
>    Control  2  2   8 12  8  7 20
>    Control  3  4   9 10  2  0 10
>    Control  4 10   8  8  5  8 14
>    Control  5 11  11 11  1  0 11
>    Control  6  1  5  15  8  9 10
>    ;

>    proc glm data=Trial;
>       class Treatment;
>       model PreY1 PostY1 FollowY1
>             PreY2 PostY2 FollowY2 = Treatment / nouni;
>       repeated Response 2 identity, Time 3;
>    run;
>=========================================

>For my problem, I just need to add another Class variable in the
>class statement. BUT, it seems in this analysis, the Response
>variables need to be Continuous data, while my Reponse are
>Ordinal data (from multiple-choice questions of sampling paper)

>So, is there any other way to do this analysis suitable for my
>data?

 
 
 

1. Special Contrast for Doubly Multivariate Design

Hello:

   I am using SPSS 6.1.3 under Windows 95.  I have a design in which six related
variables are measured under 5 different stimulus conditions.  I assume that the
six related variables are measuring a single construct and my goal is to explore
differences among the stimulus conditions on that construct.  The stimulus
factor is a within subjects factor (all subjects experience each stimulus) but
it is not temporal.  The default contrast method, polynomial, does not seem to
be appropriate for my application.  The two additional contrast methods for
within subjects factors, difference and helmert, are also not entirely
satisfactory.  Below is the syntax created for this problem using the difference
contrast.    

manova
 parmint pmskint pticint plemint pjbint
 parmjoy pmskjoy pticjoy plemjoy pjbjoy
 parmsurp pmsksurp pticsurp plemsurp pjbsurp
 parmsad pmsksad pticsad plemsad pjbsad
 parmfear pmskfear pticfear plemfear pjbfear
 parmang pmskang pticang plemang pjbang
 /wsfactor=stimulus(5)
 /contrast (stimulus)= Difference
 /measure = interest joy surprise sad fear anger
 /print = signif (multiv univ averf) transform
 /wsdesign = stimulus.

   If I wanted a more focused group of contrasts, how might I implement them
using the contrast=special option under MANOVA?  Ideally, I might like to
compare one stimulus condition to each of the others, or possibly each stimulus
condition to the grand mean.  Could anyone out there provide an example of the
use of "special" with a 5-level, within subjects factor that is not temporal?

thanks,

Chuck

2. how to detect MMU in a 68060

3. Doubly repeated measures in Proc Mixed

4. Alternate Reality!!

5. Repeated measures when time is not the repeated measure

6. Activesync 3.5 vs 3.6

7. posthoc tests with repeated measures: general linear model--repeated measures

8. Atari ST Emulator for IBM

9. Multivariate Repeated Measures

10. Multivariate vs univariate repeated measures

11. repeated measure ANOVA - multivariate?

12. Power, Sample Size Calculations in Repeated Measures Designs