Introduction to generalized linear mixed models in SPSS
Locked
Introduction to generalized linear mixed models in SPSS
 
|
I am looking for a good introduction into analyzing generalized linear mixed models in SPSS 22. The data that I need to analyze are binary (or sometimes also categorical ordered) outcomes from clinical trials comparing different treatments that are determined in subjects at multiple (correlated) time points (visits) after a baseline assessment that is used as a covariate.
Currently my questions include the following: - Is it enough to specify the repeated measures data structure and covariance type in the DATA_STRUCTURE section of the GENLINMIXED command, or do I also need to define Visit as a random effect in the RANDOM section? - Since my outcome measure is a binary response, how do I obtain estimates for the odds ratio or risk ratio (with their confidence intervals) instead of marginal means? There are tons of introductory texts on generalized linear mixed models using SAS procedure GLIMMIX, but unfortunately I haven't been able to find anything comparable for SPSS GENLINMIXED. The only text that I could find is a chapter in Marija J. Norušis' IBM SPSS Statistics 19 Advanced Statistical Procedures Companion, but unfortunately this seems to be no longer available ... Thank you, Andreas Völp |
|
|
Andreas,
I know of no such book. I recommend you learn generalized linear mixed models using GLIMMIX and then apply that knowledge to other software. With that knowledge, your empirical (Should I specify G-side random effects?) and SPSS-specific questions (How do I output odds ratios?) should become relatively easy for you to answer. Short of that, I'd recommend you seek paid for the proper analysis of data from your clinical trial, which could have a serious real-world implications. If you have built in your budget for a statistician, then you ought to use those funds to hire one who has the necessary expertise. I tend not to write to SPSS-L that often anymore because I focus on paid work, publications, etc. When possible I try to chime in, but these days it is rare. Having said that, search the archives with the keywords mixed and/or genlinmixed and you will come across *very* detailed posts by me and others discussing some of the issues you have raised. Perhaps someone else will guide you on the list but I would argue that such guidance would require much more info from you regarding your data, design, RQs, etc. Ryan > On Nov 9, 2014, at 4:27 AM, Andreas Voelp <[hidden email]> wrote: > > I am looking for a good introduction into analyzing generalized linear mixed > models in SPSS 22. The data that I need to analyze are binary (or sometimes > also categorical ordered) outcomes from clinical trials comparing different > treatments that are determined in subjects at multiple (correlated) time > points (visits) after a baseline assessment that is used as a covariate. > > Currently my questions include the following: > > - Is it enough to specify the repeated measures data structure and > covariance type in the DATA_STRUCTURE section of the GENLINMIXED command, or > do I also need to define Visit as a random effect in the RANDOM section? > > - Since my outcome measure is a binary response, how do I obtain estimates > for the odds ratio or risk ratio (with their confidence intervals) instead > of marginal means? > > There are tons of introductory texts on generalized linear mixed models > using SAS procedure GLIMMIX, but unfortunately I haven't been able to find > anything comparable for SPSS GENLINMIXED. The only text that I could find is > a chapter in Marija J. Norušis' IBM SPSS Statistics 19 Advanced Statistical > Procedures Companion, but unfortunately this seems to be no longer available > ... > > Thank you, > Andreas Völp > > > > -- > View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Introduction-to-generalized-linear-mixed-models-in-SPSS-tp5727844.html > Sent from the SPSSX Discussion mailing list archive at 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 |
Re: Introduction to generalized linear mixed models in SPSS
|
|
In reply to this post by Frank Furter
Dear Ryan:
Thanks for the advice! Trust me, I've been involved in clinical trials for almost 30 years now, and I am well aware that they have real-world implications. I do certainly not intend to fool around with complex statistics that I have only a poor understanding of ;) Actually I was just asking for some introductory reading on the implementation and syntax of GLMMs in SPSS (with focus on repeated measures in particular). Best regards, Andreas |
|
|
Perhaps you should take a look at the following source:
Heck, R. H., Thomas, S. L., & Tabata, L. N. (2013). Multilevel and longitudinal modeling with IBM SPSS. Routledge. You can "preview" parts of this on books.google.com; see: http://books.google.com/books?hl=en&lr=&id=YQSCAAAAQBAJ&oi=fnd&pg=PP1&dq=%22multilevel+and+longitudinal+modeling+with+IBM+SPSS%22&ots=5NAVxdyxNY&sig=RFytXsmuU9BRDX4myoqbvdp-fZE#v=onepage&q=%22multilevel%20and%20longitudinal%20modeling%20with%20IBM%20SPSS%22&f=false Unfortunately, it emphasizes a menu-driven approach to defining and analyzing models in its presentation but Appendix A provides the corresponding syntax (at least in my edition). This text is supposed to complement Heck and Thomas's textbook on multilevel modeling; see: http://www.amazon.com/Introduction-Multilevel-Techniques-Quantitative-Methodology/dp/1841697567/ref=sr_1_fkmr0_3?ie=UTF8&qid=1415556810&sr=8-3-fkmr0&keywords=%222nd+edition%22+%22introduction+to+multilevel+modeling+techniques%22+heck -Mike Palij New York University [hidden email] ----- Original Message ----- From: "Andreas Voelp" <[hidden email]> To: <[hidden email]> Sent: Sunday, November 09, 2014 12:32 PM Subject: Re: Introduction to generalized linear mixed models in SPSS > Dear Ryan: > > Thanks for the advice! Trust me, I've been involved in clinical trials > for > almost 30 years now, and I am well aware that they have real-world > implications. I do certainly not intend to fool around with complex > statistics that I have only a poor understanding of ;) Actually I was > just > asking for some introductory reading on the implementation and syntax > of > GLMMs in SPSS (with focus on repeated measures in particular). > > Best regards, > Andreas ===================== 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 |
|
|
In reply to this post by Frank Furter
Andreas,
Keep in mind that GEE - generalized linear models are capable of handling correlation due to repeated binary measurements on the same subject. I would strongly recommend you learn about GEE models (population-average interpretations / marginal models) before learning GLMMs, if you have not already. The question then becomes whether you want to / ought to incorporate G-side random effects due to the nature of your data and interest in evaluation of subject-specific effects. HTH. Ryan Sent from my iPhone > On Nov 9, 2014, at 4:27 AM, Andreas Voelp <[hidden email]> wrote: > > I am looking for a good introduction into analyzing generalized linear mixed > models in SPSS 22. The data that I need to analyze are binary (or sometimes > also categorical ordered) outcomes from clinical trials comparing different > treatments that are determined in subjects at multiple (correlated) time > points (visits) after a baseline assessment that is used as a covariate. > > Currently my questions include the following: > > - Is it enough to specify the repeated measures data structure and > covariance type in the DATA_STRUCTURE section of the GENLINMIXED command, or > do I also need to define Visit as a random effect in the RANDOM section? > > - Since my outcome measure is a binary response, how do I obtain estimates > for the odds ratio or risk ratio (with their confidence intervals) instead > of marginal means? > > There are tons of introductory texts on generalized linear mixed models > using SAS procedure GLIMMIX, but unfortunately I haven't been able to find > anything comparable for SPSS GENLINMIXED. The only text that I could find is > a chapter in Marija J. Norušis' IBM SPSS Statistics 19 Advanced Statistical > Procedures Companion, but unfortunately this seems to be no longer available > ... > > Thank you, > Andreas Völp > > > > -- > View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Introduction-to-generalized-linear-mixed-models-in-SPSS-tp5727844.html > Sent from the SPSSX Discussion mailing list archive at 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 |
Re: Introduction to generalized linear mixed models in SPSS
|
|
In reply to this post by Mike
Dear Mike: Thank you very much for the recommendation - meanwhile I have obtained the book. There is actually another companion to this book that is even more relevant for my particular research problem: Heck, R. H., Thomas, S. L., & Tabata, L. N. (2012). Multilevel modeling of categorical outcomes using IBM SPSS. Routledge. Aside from the literature, my ciurrent impression is that compared to SAS proc GLIMMIX and MIXED, the SPSS implementation of general and generalized linear mixed models in SPSS is still in its infancy. Unfortunately options like spatioal covariance patterns and Kenward-Roger df appear to be (still?) missing (or have I overlooked something?). Best regards Andreas |
|
|
Andrea,
My experience in this area is actually limited but from what I have read here and on the Multilevel mailing list, SPSS does appear to have some limitations relative to other programs. Depending upon the nature of the research problem, you might consider some other programs in additions to SPSS, such as structural equation programs like Mplus, and more knowledgeable people may know whether R will work as well with the advantage of being free. It could be the case that SAS may also work for you but, again, people who do these analyses routinely would be best to ask. Here's a link to the Multilevel mailing list where you can ask additional questions: http://www.nursing.manchester.ac.uk/learning/staff/mcampbell/multilevel.html -Mike Palij New York University [hidden email] ----- Original Message ----- From: "Andreas Voelp" <[hidden email]> To: <[hidden email]> Sent: Friday, November 28, 2014 5:23 AM Subject: Re: Introduction to generalized linear mixed models in SPSS > Mike Palij wrote >> Perhaps you should take a look at the following source: >> >> Heck, R. H., Thomas, S. L., & Tabata, L. N. (2013). Multilevel >> and longitudinal modeling with IBM SPSS. Routledge. > > Dear Mike: > > Thank you very much for the recommendation - meanwhile I have obtained > the > book. There is actually another companion to this book that is even > more > relevant for my particular research problem: > > Heck, R. H., Thomas, S. L., & Tabata, L. N. (2012). Multilevel > modeling of > categorical outcomes using IBM SPSS. Routledge. > > Aside from the literature, my ciurrent impression is that compared to > SAS > proc GLIMMIX and MIXED, the SPSS implementation of general and > generalized > linear mixed models in SPSS is still in its infancy. Unfortunately > options > like spatioal covariance patterns and Kenward-Roger df appear to be > (still?) > missing (or have I overlooked something?). > > Best regards > Andreas > > > > > -- > View this message in context: > http://spssx-discussion.1045642.n5.nabble.com/Introduction-to-generalized-linear-mixed-models-in-SPSS-tp5727844p5728075.html > Sent from the SPSSX Discussion mailing list archive at 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 |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |