I teach data analysis in social science courses (using SPSS) and was thinking about the inclusion of multilevel analysis in my classes. What do you think about that? Any recommendation for literature?
Multilevel Modeling of Categorical Outcomes Using IBM SPSS
By Ronald H Heck, Scott Thomas, Lynn Tabata
It takes you through a large number of examples - binary , ordinal, counts, - both through the GUI and syntax.
Can anybody refer me to some reference or video dealing with GENLINMIXED, generalized linear mixed model, in SPSS? - ResearchGate. Available from: https://www.researchgate.net/post/Can_anybody_refer_me_to_some_reference_or_video_dealing_with_GENLINMIXED_generalized_linear_mixed_model_in_SPSS [accessed Dec 5, 2016].
I concur -- Andy Field's book is phenomenal. His website, on the other hand, is bizarre (but, very entertaining... http://www.StatisticsHell.com/). Although, I don't believe he has much coverage of MLM.
What type of multilevel analysis are you planning on conducting?
My personal problem with SPSS is the interface is not all intuitive for the type of models my students want to fit (it is geared to repeated measure panel data) and it does not readily allow you to calculate the higher level residuals which I usually want to do (or at least it did not the last time that I looked)
The students have done an introductory course in SPSS in year 1, a regression modelling course in R in year 2. I teach 40 hours of contact which includes the lecture and practical hands on sessions which is based on taking the examples in Volume 1 and applying to other datasets.
The emphasis of the course is on modelling variance functions at different levels simultaenously.
Thank you all for references. I'm not interested in multilevel analysis for a specific investigation, but because I'm planning on proposing an optional discipline for social sciences graduates (sociologists, political scientists...).
Kelvyn, I know MLwiN is typically used for multilevel analysis, but I really want to use SPSS. SPSS is the software we use in all our classes and I do not have time to teach / introduce another for my students. The course will have 24 hours of contact including lecture and pratical sessions. Students already had an introductory course of regression (with SPSS). Do you think is viable?
Multilevel Analysis for Applied Research: It's Just Regression! (Methodology in the Social Sciences) by Robert Bickel PhD which is designed around using SPSS
We have a policy of exposing students to different software environments so switching software is part of how we teach.
And the Lemma course is there to be used with different software- the concepts are separate from the practice, so you could make use of whatever you want and indeed contribute back to the Centre's materials
I learnt yesterday that we are going to provide Lemma for SPSS but I do not know the timescale.
Some people like point and click - some people like syntax - and others the equations; and that is why I prefer to teach with MLwin - you know precisely what you are doing and if you do not know, you cannot do it!
Andy Field's book has a small coverage of MLM. However, the examples you find there can be articulated with the info about mixed models presented in the latest version of Joao Maroco's book (ANÁLISE ESTATISTICA com a utilização do SPSS) and with the fabulous MLM contents at UCLA website.
http://www.ats.ucla.edu/stat/spss/topics/MLM.htm
Regarding the course, I think is viable, however a solid knowledge on MRLM is an assumption for MLM courses.
I let you a link from Hayes' site with a nice intro to MLM using SPSS.
A nice and instructive text for the SPSS Mixed Procedure can be found here:
Peugh, J. L., & Enders, C. K. Using the SPSS Mixed Procedure to fit cross-sectional and longitudinal multilevel models. Educational and Psychological Measurement, 65(5), 717–741. doi:10.1177/0013164405278558
It provides examples for the two most rudimentary analyses, but if you're used to SPSS syntax it's easy to understand.
Apologies that this a question and not an answer but since we have SPSS multilevel users out there responding to Madalena's question.............
"does SPSS readily calculate the higher-level residuals?"
For my part , I want the precision-weighted estimates of the differentials from the general relationship because they are of key substantive interest. I fully understand that if I had 1000 people with repeated measures, I probably would not be so keen, but in most of the things I do want to calculate and display these differentials as they are social and insitutional contexts.
It also makes it much easier in teaching when you can display the differentials (random intercepts and slopes) and then explain the variances and covariances from there - (I do undertand in estimation that it goes the other way).
Chris Charlton at the Centre for Multilevel Modelling at Bristol has been converting our Lemma training materials for Modules 3 and 5 into SPSS form which should give you plenty of scope for an introductory class.
I have been editing them today so they should be available pretty soon. I learnt that SPSS still does not estimate the random slopes (only the variance of these) and that it cannot estimate models with complex level-1 heterogeneity for continuous predictors (it can do it for categorical predictors). However Chris has provided matrix syntax code in SPSS to calculate the slopes and he provides an IGLS algorithm for complex level-1 heterogeneity.
This is a personal thing but I still do not like the SPSS interface for mixed models!
Multilevel Modeling of Categorical Outcomes Using IBM SPSS
By Ronald H Heck, Scott Thomas, Lynn Tabata
It takes you through a large number of examples - binary , ordinal, counts, - both through the GUI and syntax.
Can anybody refer me to some reference or video dealing with GENLINMIXED, generalized linear mixed model, in SPSS? - ResearchGate. Available from: https://www.researchgate.net/post/Can_anybody_refer_me_to_some_reference_or_video_dealing_with_GENLINMIXED_generalized_linear_mixed_model_in_SPSS [accessed Dec 5, 2016].
Here is an easy and comprehensive book on Multilevel Modelling using SPSS. However, I use Mplus rather than SPSS or HLM for traditional multilevel analysis and Multilevel SEM.