Get Pairwise Comparison in GEE - Ordinal logistic model!

[sara2016]

Hello,

My study dataset includes ordinal data (a 5-point scale ranging from 1
highly unfamiliar to 5 highly familiar).I've run a GEE and chose an ‘ordinal
logistic’ model in SPSS and included:
Participant as the Subject Variable, Input mode (3 levels), and Repetition
(4 levels) as the Within subject factor, Familiarity score as the Dependent
Variable, Input mode, Repetition ,and the interaction between them as the
Predictor variables.

The thing is that it’s not possible to include/get 'Pairwise Comparison'
results for this type of data/model. So I am not sure how I can further
interpret the significant results and exactly locate the differences between
the different level of a factor!


Thanks in advance,
Sara



--
Sent from: http://spssx-discussion.1045642.n5.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

[Ryan]

Please post the syntax.

Sent from my iPhone

> On Aug 12, 2018, at 3:40 PM, sara2016 <[hidden email]> wrote:
>
> Hello,
>
> My study dataset includes ordinal data (a 5-point scale ranging from 1
> highly unfamiliar to 5 highly familiar).I've run a GEE and chose an ‘ordinal
> logistic’ model in SPSS and included:
> Participant as the Subject Variable, Input mode (3 levels), and Repetition
> (4 levels) as the Within subject factor, Familiarity score as the Dependent
> Variable, Input mode, Repetition ,and the interaction between them as the
> Predictor variables.
>
> The thing is that it’s not possible to include/get 'Pairwise Comparison'
> results for this type of data/model. So I am not sure how I can further
> interpret the significant results and exactly locate the differences between
> the different level of a factor!
>
>
> Thanks in advance,
> Sara
>
>
>
> --
> Sent from: http://spssx-discussion.1045642.n5.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

[sara2016]

Here is the Syntax

DATASET ACTIVATE DataSet1.
* Generalized Estimating Equations.
GENLIN FamilarityScore (ORDER=ASCENDING) BY Repetition InputMode
(ORDER=ASCENDING) WITH
    ProficiencyLevel
  /MODEL Repetition InputMode ProficiencyLevel Repetition*InputMode
 DISTRIBUTION=MULTINOMIAL LINK=CUMLOGIT
  /CRITERIA METHOD=FISHER(1) SCALE=1 MAXITERATIONS=100 MAXSTEPHALVING=5
PCONVERGE=1E-006(ABSOLUTE)
    SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95 LIKELIHOOD=FULL
  /REPEATED SUBJECT=Participant WITHINSUBJECT=ItemNo*Repetition*InputMode
SORT=YES
    CORRTYPE=INDEPENDENT ADJUSTCORR=YES COVB=ROBUST MAXITERATIONS=100
PCONVERGE=1e-006(ABSOLUTE)
    UPDATECORR=1
  /MISSING CLASSMISSING=EXCLUDE
  /PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION (EXPONENTIATED).


Thank you.





--
Sent from: http://spssx-discussion.1045642.n5.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

[Bruce Weaver]

Normally, I would use EMMEANS with COMPARE to get pairwise contrasts.  But as
the FM says regarding this particular model:

"Estimated marginal means are not available if the multinomial distribution
is used. If DISTRIBUTION = MULTINOMIAL on the MODEL subcommand and the
EMMEANS subcommand is specified, then EMMEANS is ignored and a warning is
issued."

I don't immediately see another straightforward approach.  

One can get GENLIN to spit out a vector of regression coefficients and a
covariance matrix for the parameters.  So with some application of matrix
algebra, one should be able to make whatever comparisons they wish.  Several
years ago, I wrote a macro to do that (when the boss was asking awkward
questions).  It works well for ordinary (binary) logistic regression, but I
never considered whether it would work with a multinomial outcome variable.
I would not be surprised if it needed some tweaking.  You can find that
macro (!MLEcomp) with an example here:

 
https://sites.google.com/a/lakeheadu.ca/bweaver/Home/statistics/spss/olscomp_mlecomp

If this is a one-time job and elegance is not high on the priority list,
another clunky but less complicated way might be to estimate the model
multiple times, changing the reference category for the variable(s) of
interest each time.  By doing this, I think you should be able to get all of
the comparisons of interest.  (If I was going to try this, I'd use OMS to
write the parameter estimates out to another data set.)  

But with any luck, someone else will come through with a much simpler
approach.  

HTH.



sara2016 wrote

> Here is the Syntax
>
> DATASET ACTIVATE DataSet1.
> * Generalized Estimating Equations.
> GENLIN FamilarityScore (ORDER=ASCENDING) BY Repetition InputMode
> (ORDER=ASCENDING) WITH
>     ProficiencyLevel
>   /MODEL Repetition InputMode ProficiencyLevel Repetition*InputMode
>  DISTRIBUTION=MULTINOMIAL LINK=CUMLOGIT
>   /CRITERIA METHOD=FISHER(1) SCALE=1 MAXITERATIONS=100 MAXSTEPHALVING=5
> PCONVERGE=1E-006(ABSOLUTE)
>     SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95 LIKELIHOOD=FULL
>   /REPEATED SUBJECT=Participant WITHINSUBJECT=ItemNo*Repetition*InputMode
> SORT=YES
>     CORRTYPE=INDEPENDENT ADJUSTCORR=YES COVB=ROBUST MAXITERATIONS=100
> PCONVERGE=1e-006(ABSOLUTE)
>     UPDATECORR=1
>   /MISSING CLASSMISSING=EXCLUDE
>   /PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION (EXPONENTIATED).
>
>
> Thank you.
>
>
>
>
>
> --
> Sent from: http://spssx-discussion.1045642.n5.nabble.com/
>
> =====================
> To manage your subscription to SPSSX-L, send a message to

> LISTSERV@.UGA

>  (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
[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.

--
Sent from: http://spssx-discussion.1045642.n5.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

[Ryan]

As an aside, I have concerns about the “ItemNo*Repetition*InputMode” within-subject and “independent” covariance structure specifications.  

> On Aug 14, 2018, at 2:43 PM, Bruce Weaver <[hidden email]> wrote:
>
> Normally, I would use EMMEANS with COMPARE to get pairwise contrasts.  But as
> the FM says regarding this particular model:
>
> "Estimated marginal means are not available if the multinomial distribution
> is used. If DISTRIBUTION = MULTINOMIAL on the MODEL subcommand and the
> EMMEANS subcommand is specified, then EMMEANS is ignored and a warning is
> issued."
>
> I don't immediately see another straightforward approach.  
>
> One can get GENLIN to spit out a vector of regression coefficients and a
> covariance matrix for the parameters.  So with some application of matrix
> algebra, one should be able to make whatever comparisons they wish.  Several
> years ago, I wrote a macro to do that (when the boss was asking awkward
> questions).  It works well for ordinary (binary) logistic regression, but I
> never considered whether it would work with a multinomial outcome variable.
> I would not be surprised if it needed some tweaking.  You can find that
> macro (!MLEcomp) with an example here:
>
>
> https://sites.google.com/a/lakeheadu.ca/bweaver/Home/statistics/spss/olscomp_mlecomp
>
> If this is a one-time job and elegance is not high on the priority list,
> another clunky but less complicated way might be to estimate the model
> multiple times, changing the reference category for the variable(s) of
> interest each time.  By doing this, I think you should be able to get all of
> the comparisons of interest.  (If I was going to try this, I'd use OMS to
> write the parameter estimates out to another data set.)  
>
> But with any luck, someone else will come through with a much simpler
> approach.  
>
> HTH.
>
>
>
> sara2016 wrote
>> Here is the Syntax
>>
>> DATASET ACTIVATE DataSet1.
>> * Generalized Estimating Equations.
>> GENLIN FamilarityScore (ORDER=ASCENDING) BY Repetition InputMode
>> (ORDER=ASCENDING) WITH
>>    ProficiencyLevel
>>  /MODEL Repetition InputMode ProficiencyLevel Repetition*InputMode
>> DISTRIBUTION=MULTINOMIAL LINK=CUMLOGIT
>>  /CRITERIA METHOD=FISHER(1) SCALE=1 MAXITERATIONS=100 MAXSTEPHALVING=5
>> PCONVERGE=1E-006(ABSOLUTE)
>>    SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95 LIKELIHOOD=FULL
>>  /REPEATED SUBJECT=Participant WITHINSUBJECT=ItemNo*Repetition*InputMode
>> SORT=YES
>>    CORRTYPE=INDEPENDENT ADJUSTCORR=YES COVB=ROBUST MAXITERATIONS=100
>> PCONVERGE=1e-006(ABSOLUTE)
>>    UPDATECORR=1
>>  /MISSING CLASSMISSING=EXCLUDE
>>  /PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION (EXPONENTIATED).
>>
>>
>> Thank you.
>>
>>
>>
>>
>>
>> --
>> Sent from: http://spssx-discussion.1045642.n5.nabble.com/
>>
>> =====================
>> To manage your subscription to SPSSX-L, send a message to
>
>> LISTSERV@.UGA
>
>> (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
> [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.
>
> --
> Sent from: http://spssx-discussion.1045642.n5.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]

Good point, Ryan.  I had not noticed that.


Ryan Black wrote

> As an aside, I have concerns about the “ItemNo*Repetition*InputMode”
> within-subject and “independent” covariance structure specifications.  
>
>
>> sara2016 wrote
>>> Here is the Syntax
>>>
>>> DATASET ACTIVATE DataSet1.
>>> * Generalized Estimating Equations.
>>> GENLIN FamilarityScore (ORDER=ASCENDING) BY Repetition InputMode
>>> (ORDER=ASCENDING) WITH
>>>    ProficiencyLevel
>>>  /MODEL Repetition InputMode ProficiencyLevel Repetition*InputMode
>>> DISTRIBUTION=MULTINOMIAL LINK=CUMLOGIT
>>>  /CRITERIA METHOD=FISHER(1) SCALE=1 MAXITERATIONS=100 MAXSTEPHALVING=5
>>> PCONVERGE=1E-006(ABSOLUTE)
>>>    SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95 LIKELIHOOD=FULL
>>>  /REPEATED SUBJECT=Participant WITHINSUBJECT=ItemNo*Repetition*InputMode
>>> SORT=YES
>>>    CORRTYPE=INDEPENDENT ADJUSTCORR=YES COVB=ROBUST MAXITERATIONS=100
>>> PCONVERGE=1e-006(ABSOLUTE)
>>>    UPDATECORR=1
>>>  /MISSING CLASSMISSING=EXCLUDE
>>>  /PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION (EXPONENTIATED).
>>>

-----
--
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.

--
Sent from: http://spssx-discussion.1045642.n5.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

[sara2016]

Thanks a lot for your reply.

That's actually how I reported my results/comparisons so far -  based on the
Parameter Estimates. But not sure if that's enough and justified!


Best
Sara



--
Sent from: http://spssx-discussion.1045642.n5.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

[sara2016]

Thank you.



--
Sent from: http://spssx-discussion.1045642.n5.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