Mixed model and time variable query

[Paola Chivers-2]

HI



I am using my bible:  Chpt 5 Singer & Willet and they suggest that it is
more appropriate to use the most accurate source of time (in my case this is
'age' which is each individual's age at assessment), followed by 'Time' (
average age at each assessment 0,1,2,3,6,8,10.5,14) and lastly 'timeN'
(consecutive time scale 1-8).



So...



MODEL SINGER AND WILLET FOR TREATING TIME MORE FLEXIBLY WHERE CAGEGRP IS THE
EQUIVALENT TO MY 'Time'.  However they also run the 'age' model and present
the results but don't give us the syntax - and 'age' model is better, hence
the rationale for using time more flexibly.  So I have assumed that it would
be most appropriate to use 'age'.



mixed piat with cagegrp

  /print=solution

  /method=ml

  /fixed=intercept cagegrp

  /random=intercept cagegrp | subject(id) covtype(un).



Copying this syntax I use the following ... note that a repeated measures is
not included in the model!!!  And It works.



mixed bmi with age

  /print=solution

  /method=ml

  /fixed=intercept age

  /random=intercept age | subject(studyno) covtype(un).



However, if I then try to include the repeated measure - things change.  I
don't really understand why they are not using the repeated part.



In this model I use 'age' as the covariate, and 'time' the repeated measure.
Results work.



MIXED

  BMI  WITH Age

  /CRITERIA = CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR

 (0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE

 (0.000001, ABSOLUTE)

  /FIXED = Age  | SSTYPE(3)

  /METHOD = REML

  /REPEATED = Time | SUBJECT(STUDYNO) COVTYPE(UN) .



In this model I use 'age' as both the covariate, and the repeated measure
and get the following error message.




MIXED

  BMI  WITH Age

  /CRITERIA = CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR

 (0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE

 (0.000001, ABSOLUTE)

  /FIXED = Age  | SSTYPE(3)

  /METHOD = REML

  /REPEATED = Age | SUBJECT(STUDYNO) COVTYPE(UN) .



Warnings




Insufficient memory to estimate the model parameters.


This command is not executed.





What is my critical mistake this time???





Regards,

Paola



"Ours has become a time-poor society, fatigued by non-physical demands and
trying to compartmentalize daily living tasks.  It is small wonder that
physical activity is discarded in this environment" p126 (Steinbeck, 2001)



P Please consider the environment before printing this email.

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

[Swank, Paul R]

Do you wish to do a random coefficients model or a repeated measures
model? I can understand putting time in as a repeated measure with age
as a random coefficient if you think that the random coefficient model
with not account for all the variation in the DV but putting age as a
repeated factor will usually give too large a matrix. The variance
covariance structure for random coefficients models and repeated models
are different. With a random intercept and slope you have a 2 x 2
matrix. With four repeated observations you have a 4 x 4 for each
subject. But if you have too many levels of age, then specifying age as
a repeated factor is likely to produce the problems you are seeing.

Paul R. Swank, Ph.D.
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center - Houston


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Paola Chivers
Sent: Thursday, May 08, 2008 9:13 PM
To: [hidden email]
Subject: Mixed model and time variable query

HI



I am using my bible:  Chpt 5 Singer & Willet and they suggest that it is
more appropriate to use the most accurate source of time (in my case
this is
'age' which is each individual's age at assessment), followed by 'Time'
(
average age at each assessment 0,1,2,3,6,8,10.5,14) and lastly 'timeN'
(consecutive time scale 1-8).



So...



MODEL SINGER AND WILLET FOR TREATING TIME MORE FLEXIBLY WHERE CAGEGRP IS
THE
EQUIVALENT TO MY 'Time'.  However they also run the 'age' model and
present
the results but don't give us the syntax - and 'age' model is better,
hence
the rationale for using time more flexibly.  So I have assumed that it
would
be most appropriate to use 'age'.



mixed piat with cagegrp

  /print=solution

  /method=ml

  /fixed=intercept cagegrp

  /random=intercept cagegrp | subject(id) covtype(un).



Copying this syntax I use the following ... note that a repeated
measures is
not included in the model!!!  And It works.



mixed bmi with age

  /print=solution

  /method=ml

  /fixed=intercept age

  /random=intercept age | subject(studyno) covtype(un).



However, if I then try to include the repeated measure - things change.
I
don't really understand why they are not using the repeated part.



In this model I use 'age' as the covariate, and 'time' the repeated
measure.
Results work.



MIXED

  BMI  WITH Age

  /CRITERIA = CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR

 (0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE)
PCONVERGE

 (0.000001, ABSOLUTE)

  /FIXED = Age  | SSTYPE(3)

  /METHOD = REML

  /REPEATED = Time | SUBJECT(STUDYNO) COVTYPE(UN) .



In this model I use 'age' as both the covariate, and the repeated
measure
and get the following error message.




MIXED

  BMI  WITH Age

  /CRITERIA = CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR

 (0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE)
PCONVERGE

 (0.000001, ABSOLUTE)

  /FIXED = Age  | SSTYPE(3)

  /METHOD = REML

  /REPEATED = Age | SUBJECT(STUDYNO) COVTYPE(UN) .



Warnings




Insufficient memory to estimate the model parameters.


This command is not executed.





What is my critical mistake this time???





Regards,

Paola



"Ours has become a time-poor society, fatigued by non-physical demands
and
trying to compartmentalize daily living tasks.  It is small wonder that
physical activity is discarded in this environment" p126 (Steinbeck,
2001)



P Please consider the environment before printing this email.

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

[Dale Glaser]

Hi all......I am working on a study with a intent to mirror the analysis of Plummer and Owen (2001) in Anest Analg "Learning Endotracheal intubation...".....they had a repeated measures design with a binary outcome and conducted a random effects logistic regression model by using the xtlogit procedure of Stata...........does anyone know if I would be able to fit a similar model using the GEE option in SPSS with the logit link function?..or is this type of analysis something only Stata (or R) can do?

  thank you very much...Dale Glaser


Dale Glaser, Ph.D.
Principal--Glaser Consulting
Lecturer/Adjunct Faculty--SDSU/USD/AIU
President, San Diego Chapter of
American Statistical Association
3115 4th Avenue
San Diego, CA 92103
phone: [phone]
fax: [phone]
email: [hidden email]
website: www.glaserconsult.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