Multilevel Poisson regression using IBM SPSS (March 2020)
HTML-код
- Опубликовано: 21 авг 2024
- The aim of this demonstration is to show you some of the basics of performing multilevel Poisson regression through IBM SPSS. You can download a copy of the data here: drive.google.c...
Download a copy of the Powerpoint here: drive.google.c...
To learn more about Heck et al. (2012) book go here: www.routledge....
Thank you for the detailed go-through :)
Thank you so much Professor
Stay safe🙏🏻
Awesome. Thanks!
Thanks so much
Great vid. Why is the lowSES set as a random variable, but the others are not? thx
Hi there. It was just to demonstrate how you can specify a random slope, and nothing substantive. But I will say that you need to be careful with specification of random slopes for a lot of predictors. As the model becomes increasingly complex you run the risk of greater problems with model convergence. Cheers!
@@mikecrowson2462 Got it, Thanks! BTW, is there a difference using the 'GLM Mixed' option versus the 'Generalized Estimating Equation' option in the SPSS menu when running a survival/repeated measure model? The setup seems the same, with the GEE perhaps a bit more control features? Thanks for the great videos!
@@molloarden8938 The generalized estimating equations option approaches estimation using a quasi-likelihood function. So, it's mainly used for estimating model parameters. It is handy if that is your goal, but it becomes trickier when it comes to trying to evaluate the overall fit of a model. But is IS easier to implement than the mixed option, as there is no modeling of random coefficients. I don't do a lot with GEE (which by the way might be considered an alternative to panel regression), but Daniel McNeish has written on the use of this approach in the context of cross-sectional data [he does a fantastic job of showing you the similarities and differences between panel regression, GEE, and multilevel regression]. But it works the same theoretically, since the 'clusters' in repeated measures are at the individual level (whereas the 'clusters' with cross sectional data might be organizations, classrooms, etc). I found this article that you might check out: www.tandfonline.com/doi/full/10.1080/00273171.2016.1167008
and another: www.researchgate.net/publication/264627635_Modeling_Sparsely_Clustered_Data_Design-Based_Model-Based_and_Single-Level_Methods
You should also check out: asu.pure.elsevier.com/en/publications/on-the-unnecessary-ubiquity-of-hierarchical-linear-modeling
@@mikecrowson2462 Much appreciated the links, given the paucity of research on comparative value of mixed-level versus GEE models. Here's a link to mixed-level study I did years ago before GEE options in SPSS. Kudos to you for the lucid videos! eric.ed.gov/?id=ED508971
Hi, does anyone know how would you specify the same model in R with GLMER? I'm interested in knowing how to introduce the level 2 predictors.
Hey! Could you help me to understand how to download the model of results? I mean that I could make a screenshot but maybe there is one more way to download it accurately...thanks!!