SPSSX Discussion

spss 19 genlinmixed

Classic

List

Threaded

4 messages Options

marc glassman

spss 19 genlinmixed

Recently, I have been trying to use the new GENLINMIXED procedure in
IBM/SPSS 19. I confess that I find the Model Viewer unwieldy to use as a
function of its multiple page "views" rather than the more conventional
output format used by the MIXED and GENLIN procedures, but, more
importantly, I have had difficulty printing or saving the output from the
procedure.

It's possible that these difficulties are specific to my computer, but I'd
be interested in knowing whether others have similar problems and, if so,
what steps they've taken to resolve them.

Thanks . . .

Marc Glassman

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

Michael N. Mitchell

Re: spss 19 genlinmixed

Dear Marc

I was struggling with the same issues. I was completely frustrated by the how much
pointing and clicking I had to do to see standard and generally accepted statistical
ouptut. As you note, once you click enough times to get the output, it is displayed in a
completely fragmented way. I could not find any way to solve this, and ultimately switched
to using another tool for the analysis. I fret that this might be a prototype for how
other modeling procedures might be implemented to fit in with the new SPSS modeling
methodologies.

Michael N. Mitchell
Data Management Using Stata - http://www.stata.com/bookstore/dmus.html
A Visual Guide to Stata Graphics - http://www.stata.com/bookstore/vgsg.html
Stata tidbit of the week - http://www.MichaelNormanMitchell.com

On 2011-01-31 4.23 PM, marc glassman wrote:

> Recently, I have been trying to use the new GENLINMIXED procedure in
> IBM/SPSS 19. I confess that I find the Model Viewer unwieldy to use as a
> function of its multiple page "views" rather than the more conventional
> output format used by the MIXED and GENLIN procedures, but, more
> importantly, I have had difficulty printing or saving the output from the
> procedure.
>
> It's possible that these difficulties are specific to my computer, but I'd
> be interested in knowing whether others have similar problems and, if so,
> what steps they've taken to resolve them.
>
> Thanks . . .
>
>
> Marc Glassman
>
> =====================
> 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

Re: spss 19 genlinmixed

I am a fan of generalized linear mixed models (GLMMs), given their
flexibility to handle various designs (nested, crossed, repeated
measures etc.), response distributions (binary/binomial, multinomial,
poisson, gaussian etc.), and link functions (e.g. identity, logit, log
etc.). As such, I was delighted to hear that a GLMM procedure had been
Incorporated into v19. Much to my dismay, however, the GENLINMIXED
procedure fell very short of my expectations. Here are a few issues
that are deal breakers for me:

(1) Adaptive gaussian approximation methods are not offered
(2) A flexible contrast subcommand (e.g., TEST in MIXED) is not offered
(3) Combing through the output is tedious

These three key components are all offered in the GLIMMIX procedure in
SAS. Until SPSS catches up, I, among many others I suspect, will
continue to utilize software such as SAS instead.

I'd be happy to be corrected on any one of my concerns. Further
thoughts on this matter are welcome.

Ryan

On Wed, Feb 2, 2011 at 6:07 PM, Michael N. Mitchell
<[hidden email]> wrote:

> Dear Marc
>
> I was struggling with the same issues. I was completely frustrated by the
> how much
> pointing and clicking I had to do to see standard and generally accepted
> statistical
> ouptut. As you note, once you click enough times to get the output, it is
> displayed in a
> completely fragmented way. I could not find any way to solve this, and
> ultimately switched
> to using another tool for the analysis. I fret that this might be a
> prototype for how
> other modeling procedures might be implemented to fit in with the new SPSS
> modeling
> methodologies.
>
> Michael N. Mitchell
> Data Management Using Stata - http://www.stata.com/bookstore/dmus.html
> A Visual Guide to Stata Graphics - http://www.stata.com/bookstore/vgsg.html
> Stata tidbit of the week - http://www.MichaelNormanMitchell.com
>
>
>
> On 2011-01-31 4.23 PM, marc glassman wrote:
>>
>> Recently, I have been trying to use the new GENLINMIXED procedure in
>> IBM/SPSS 19. I confess that I find the Model Viewer unwieldy to use as a
>> function of its multiple page "views" rather than the more conventional
>> output format used by the MIXED and GENLIN procedures, but, more
>> importantly, I have had difficulty printing or saving the output from the
>> procedure.
>>
>> It's possible that these difficulties are specific to my computer, but I'd
>> be interested in knowing whether others have similar problems and, if so,
>> what steps they've taken to resolve them.
>>
>> Thanks . . .
>>
>>
>> Marc Glassman
>>
>> =====================
>> 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
>

Michael N. Mitchell

Re: spss 19 genlinmixed

Dear Ryan

You raise some very good points beyond the issue of the tedious output management. I
would add that the syntax is the most verbose I have ever seen. I am also troubled that,
by default, I believe that for a 0/1 outcome treats the 0 as the event of interest.

I have had trouble finding technical documentation regarding the estimation methods
used by GENLINMIXED, but it appears that it is using PQL (penalized quasi likelihood). As
described in http://www2.sas.com/proceedings/forum2007/177-2007.pdf, PQL methods can
produce biased estimates, as compared to numerical integration methods as used by SAS with
GLIMMIX or with Stata with something like xtmelogit.

Michael N. Mitchell
Data Management Using Stata - http://www.stata.com/bookstore/dmus.html
A Visual Guide to Stata Graphics - http://www.stata.com/bookstore/vgsg.html
Stata tidbit of the week - http://www.MichaelNormanMitchell.com

On 2011-02-02 6.31 PM, R B wrote:

> I am a fan of generalized linear mixed models (GLMMs), given their
> flexibility to handle various designs (nested, crossed, repeated
> measures etc.), response distributions (binary/binomial, multinomial,
> poisson, gaussian etc.), and link functions (e.g. identity, logit, log
> etc.). As such, I was delighted to hear that a GLMM procedure had been
> Incorporated into v19. Much to my dismay, however, the GENLINMIXED
> procedure fell very short of my expectations. Here are a few issues
> that are deal breakers for me:
>
> (1) Adaptive gaussian approximation methods are not offered
> (2) A flexible contrast subcommand (e.g., TEST in MIXED) is not offered
> (3) Combing through the output is tedious
>
> These three key components are all offered in the GLIMMIX procedure in
> SAS. Until SPSS catches up, I, among many others I suspect, will
> continue to utilize software such as SAS instead.
>
> I'd be happy to be corrected on any one of my concerns. Further
> thoughts on this matter are welcome.
>
> Ryan
>
> On Wed, Feb 2, 2011 at 6:07 PM, Michael N. Mitchell
> <[hidden email]> wrote:
>> Dear Marc
>>
>> I was struggling with the same issues. I was completely frustrated by the
>> how much
>> pointing and clicking I had to do to see standard and generally accepted
>> statistical
>> ouptut. As you note, once you click enough times to get the output, it is
>> displayed in a
>> completely fragmented way. I could not find any way to solve this, and
>> ultimately switched
>> to using another tool for the analysis. I fret that this might be a
>> prototype for how
>> other modeling procedures might be implemented to fit in with the new SPSS
>> modeling
>> methodologies.
>>
>> Michael N. Mitchell
>> Data Management Using Stata - http://www.stata.com/bookstore/dmus.html
>> A Visual Guide to Stata Graphics - http://www.stata.com/bookstore/vgsg.html
>> Stata tidbit of the week - http://www.MichaelNormanMitchell.com
>>
>>
>>
>> On 2011-01-31 4.23 PM, marc glassman wrote:
>>>
>>> Recently, I have been trying to use the new GENLINMIXED procedure in
>>> IBM/SPSS 19. I confess that I find the Model Viewer unwieldy to use as a
>>> function of its multiple page "views" rather than the more conventional
>>> output format used by the MIXED and GENLIN procedures, but, more
>>> importantly, I have had difficulty printing or saving the output from the
>>> procedure.
>>>
>>> It's possible that these difficulties are specific to my computer, but I'd
>>> be interested in knowing whether others have similar problems and, if so,
>>> what steps they've taken to resolve them.
>>>
>>> Thanks . . .
>>>
>>>
>>> Marc Glassman
>>>
>>> =====================
>>> 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
>>
>
> =====================
> 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