DF in multilevel models

[Baglioni, Tony]

All,

 

I am running a simple multilevel model (fixed effects only) for a
colleague.  The results are fairly easy to interpret but I am unclear
how to explain the degrees of freedom reported by SPSS in the Estimates
of Fixed Effects table.  The data consists of 48 individuals observed on
four occasions.  The df have values (depending on the number of fixed
effects) in the 144 to 153 range.  The part I do not understand is why
the df are reported with decimals; that is, 144.32 or 146.81.  I cannot
find a clear explanation in the books I have (Singer & Willet or Bryk &
Raudenbush or Snijders and Bosker) of how DF are obtained.

 

Can anyone offer a simple explanation?

 

Thanks in advance,

Tony

 

A J Baglioni jr, PhD

Assistant Professor

McIntire School of Commerce

University of Virginia

Charlottesville VA 22904-4173

434.924.4961

 

[peter link]

Tony -

I can give you a simple explanation, but it might not be all of that
helpful, I'm afraid.  The denominator degrees of freedom are estimated using
the Satterthwaite method.  To keep this simple, it basically takes into
account the total number of observations that you have for each variable.
One variable may have larger (or smaller) df due, at least in part, to less
(or more) missing data.

To learn more, you should see the algorithm section for MIXED on the tech
support site.  It shows how the df are estimated (with formulas).  This will
probably only be helpful if you know some Mathematical Statistics and some
Linear Algebra.

This is the website

http://support.spss.com/Tech/

If you need to enter user/pass, enter guest for both.  Click "Statistics
Documentation" and then "Algorithms" and then "Mixed".

Peter Link
VA San Diego Healthcare System

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
Baglioni, Tony
Sent: Monday, June 25, 2007 6:28 AM
To: [hidden email]
Subject: DF in multilevel models


All,



I am running a simple multilevel model (fixed effects only) for a
colleague.  The results are fairly easy to interpret but I am unclear
how to explain the degrees of freedom reported by SPSS in the Estimates
of Fixed Effects table.  The data consists of 48 individuals observed on
four occasions.  The df have values (depending on the number of fixed
effects) in the 144 to 153 range.  The part I do not understand is why
the df are reported with decimals; that is, 144.32 or 146.81.  I cannot
find a clear explanation in the books I have (Singer & Willet or Bryk &
Raudenbush or Snijders and Bosker) of how DF are obtained.



Can anyone offer a simple explanation?



Thanks in advance,

Tony



A J Baglioni jr, PhD

Assistant Professor

McIntire School of Commerce

University of Virginia

Charlottesville VA 22904-4173

434.924.4961

[Pirritano, Matthew]

Hello all,

I have a multilevel model that I cannot get to run. It involves the use of David Kenny's actor partner interdependence model to run an analysis where level one only has two levels.

The data are organized in the following format. What David Kenny and colleagues refer to as the pairwise data format. Each individual has both their own data (labeled actor) and the other member of the dyad's data (labeled partner):

dyad    sex     dv      actor    partner
1        1      5          8       6
1       -1      4          6       8
2        1      6          4       2
2       -1      3          2       4
3        1      2          1       9
3       -1      8          9       1


In my model I am predicting an individual's level of stress from their own coping style and their partner's coping style.

My syntax is:

mixed
        apd with act_aaacop part_aaacop sex
        /fixed = act_aaacop part_aaacop sex act_aaacop*sex part_aaacop*sex | noint
        /print = solution testcov
        /repeated = sex | subject(dyad) covtype(csh).

The error message I'm getting is:

The levels of the repeated effect are not different for each observation within a repeated subject.
This command is not executed.

Any help would be much appreciated. Please let me know if I need to further clarify the question.

Thanks,
Matt

Matthew Pirritano, Ph.D.
Assistant Professor of Psychology
Smith Hall 116C
Chapman University
Department of Psychology
One University Drive
Orange, CA 92866
Telephone [phone]
FAX [phone]

[peter link]

Matthew -

I went through and repeated the syntax that you gave below, and I did not
receive any error message; it worked fine.  I'm not sure what the problem
may be.  From the error that you post, I would guess that the values of sex
are not always different within each dyad.  I would go back and double check
that these values are different for each partner that makes up the dyad.  At
least that would be a good place to start.

Peter

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
Pirritano, Matthew
Sent: Monday, June 25, 2007 9:42 AM
To: [hidden email]
Subject: multilevel model help


Hello all,

I have a multilevel model that I cannot get to run. It involves the use of
David Kenny's actor partner interdependence model to run an analysis where
level one only has two levels.

The data are organized in the following format. What David Kenny and
colleagues refer to as the pairwise data format. Each individual has both
their own data (labeled actor) and the other member of the dyad's data
(labeled partner):

dyad    sex     dv      actor    partner
1        1      5          8       6
1       -1      4          6       8
2        1      6          4       2
2       -1      3          2       4
3        1      2          1       9
3       -1      8          9       1


In my model I am predicting an individual's level of stress from their own
coping style and their partner's coping style.

My syntax is:

mixed
        apd with act_aaacop part_aaacop sex
        /fixed = act_aaacop part_aaacop sex act_aaacop*sex part_aaacop*sex |
noint
        /print = solution testcov
        /repeated = sex | subject(dyad) covtype(csh).

The error message I'm getting is:

The levels of the repeated effect are not different for each observation
within a repeated subject.
This command is not executed.

Any help would be much appreciated. Please let me know if I need to further
clarify the question.

Thanks,
Matt

Matthew Pirritano, Ph.D.
Assistant Professor of Psychology
Smith Hall 116C
Chapman University
Department of Psychology
One University Drive
Orange, CA 92866
Telephone [phone]
FAX [phone]

[Ornelas, Fermin]

There is an excellent paper by Sharon Lhor from ASU illustrating how to
keep track of the degrees of freedom.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
peter link
Sent: Monday, June 25, 2007 9:24 AM
To: [hidden email]
Subject: Re: DF in multilevel models

Tony -

I can give you a simple explanation, but it might not be all of that
helpful, I'm afraid.  The denominator degrees of freedom are estimated
using
the Satterthwaite method.  To keep this simple, it basically takes into
account the total number of observations that you have for each
variable.
One variable may have larger (or smaller) df due, at least in part, to
less
(or more) missing data.

To learn more, you should see the algorithm section for MIXED on the
tech
support site.  It shows how the df are estimated (with formulas).  This
will
probably only be helpful if you know some Mathematical Statistics and
some
Linear Algebra.

This is the website

http://support.spss.com/Tech/

If you need to enter user/pass, enter guest for both.  Click "Statistics
Documentation" and then "Algorithms" and then "Mixed".

Peter Link
VA San Diego Healthcare System

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
Baglioni, Tony
Sent: Monday, June 25, 2007 6:28 AM
To: [hidden email]
Subject: DF in multilevel models


All,



I am running a simple multilevel model (fixed effects only) for a
colleague.  The results are fairly easy to interpret but I am unclear
how to explain the degrees of freedom reported by SPSS in the Estimates
of Fixed Effects table.  The data consists of 48 individuals observed on
four occasions.  The df have values (depending on the number of fixed
effects) in the 144 to 153 range.  The part I do not understand is why
the df are reported with decimals; that is, 144.32 or 146.81.  I cannot
find a clear explanation in the books I have (Singer & Willet or Bryk &
Raudenbush or Snijders and Bosker) of how DF are obtained.



Can anyone offer a simple explanation?



Thanks in advance,

Tony



A J Baglioni jr, PhD

Assistant Professor

McIntire School of Commerce

University of Virginia

Charlottesville VA 22904-4173

434.924.4961

NOTICE: This e-mail (and any attachments) may contain PRIVILEGED OR
CONFIDENTIAL information and is intended only for the use of the
specific individual(s) to whom it is addressed.  It may contain
information that is privileged and confidential under state and federal
law.  This information may be used or disclosed only in accordance with
law, and you may be subject to penalties under law for improper use or
further disclosure of the information in this e-mail and its
attachments. If you have received this e-mail in error, please
immediately notify the person named above by reply e-mail, and then
delete the original e-mail.  Thank you.