Multilevel regression using Stata: Modeling two-level data (Dec. 2019)

Рет қаралды 41,009

4 жыл бұрын

This video provides a walk through of multilevel regression modeling using Stata, where the data falls at two-levels (in this case, students at Level 1 nested within schools at Level 2). I walk through a general "build up" strategy where I test the following models: (1) null model (no predictors), (2) level 1 predictors added, (3) level 2 predictors added, (4) random slopes added, (5) test of cross-level interaction.
A copy of the data can be downloaded here:
drive.google.com/file/d/187DJ...
A copy of the text file referenced in the video can be downloaded here drive.google.com/file/d/1CQT-...
A pdf of the outputs generated can be obtained here drive.google.com/file/d/1I6rw...
For more videos on multilevel regression in Stata and other programs, go here:
sites.google.com/view/statist...

Пікірлер: 35

@mastorbatrix 4 жыл бұрын

Thank you! You cannot imagine how helpful this is when writing my thesis during this corona lockdown!

@user-fp2rd1nb3y Жыл бұрын

Thank you!! It's very helpful!

@billestrada2905 3 жыл бұрын

Thank you! it´s fantastic!

@MahmoudElkasabi 4 жыл бұрын

Thanks Mike for such useful illustrations. I'm interested in the same application but on survey data where survey weights should be incorporated , and weight re-scaling might be needed. Did you do any videos about this type of analysis?

@JasonBurtt 4 жыл бұрын

Do you need to allow the slope for SES to randomly vary (in this case varying independently from intercept) in order to use it in an interaction term?

@keremmorgul367 3 жыл бұрын

Great video, very clear explanations. But I'd like to note that the use of ICC as an indicator of whether the researcher should use multilevel modeling is disputed. ICC, after all, is a function of between-group variation in means. Even when there is little variation in group means, the relationships of interest (e.g. the one between s.e.s. and math score) may vary across the groups.

@Alsaad1985 4 жыл бұрын

Thank you for this useful video I have a question. I run several mixed effect models. In my results, ICC for the level 1 model is higher than the null model. And ICC for the level 2 model is less that ICC for the level 1. What does this mean?

@Lara-zo8gn Жыл бұрын

Very helpful, thank you! One question: Do I include control variables like age or gender already in the Nullmodel?

@Alsaad1985 4 жыл бұрын

Another question Is the mixed effects model usefull when I have a daily time serise variable (level 1) and country cultural effects ( level 2)?

@asayealamneh4085 4 жыл бұрын

Thank you very much for your kindly feedback, Mr. Mike Crowson Thank you very much for your kindly feedback, Ph.D. Mike Crowson, How can we check influential and outlier observations in the multilevel binary logistic regression model in Stata?

@glossina3 3 жыл бұрын

thank you very much for your video, I just want to know how can I get the fixed effect test (F-TEST IN SPSS) in stata

@amanrastogi5184 4 жыл бұрын

Sir, can we have a random slope of a 2nd level variable in a 2 level model.

@AS-ey7ge Жыл бұрын

Great video. It was very helpful. Does the multilevel regression have assumptions like linear regression? How can they be tested in Stata? Do you have any literature recommendations?

@Hiwaradan Жыл бұрын

Thanks for this great video. I have a 2 level sem modetating mediation model. Can I do the the cfa and hypotheses test with stata?

@texkaltlipoka666 3 жыл бұрын

Thank you! Very useful video. Samurais function rules!

@mikecrowson2462 3 жыл бұрын

You are very welcome, Rodrigo!

@semagn2127 3 жыл бұрын

Thanks Dr. could you explain the basic concepts like random effect, fixed effect, linear mixed effect model, generalized linear mixed effect, Generalized estimated equation...their difference in concept .just one video it will help us a lot especially for beginners

@kchou3153 3 жыл бұрын

how can I get standardized coefficients (beta) in multilevel regression ?

@JasonBurtt 4 жыл бұрын

p.s. thanks for this awesome video

@Alejandroperezpita 3 жыл бұрын

Hello Mike, first of all thank you so much for your videos! What I'm finding is that I can't run the "lrtest" command to compare my models because the number of observations is different for each of them, depending on the missing data they are affected by, Do you have any video on how to deal with the missing data in order to properly carry out multi-level modeling?

@mikecrowson2462 3 жыл бұрын

Hi there. Unfortunately, I have yet to find a way around this problem, apart from ensuring there is no difference in the number of missing cases between the models. I suppose you could listwise delete cases that are likely to create this problem, or use some type of imputation. Believe me, if I find something that works, I'll it up! Thanks for your comment and question!

@Alejandroperezpita 3 жыл бұрын

@@mikecrowson2462 Thanks for your reply. I have been trying multiple imputation as a means to solve the problem, but the knowledge required to carry it out in such a large dataset (I am using TIMSS 2015) is far beyond me. Listwise deletion would greatly decrease the size of my sample. Do you think doing "lrtest m1 m2, force" is acceptable? Or could I just not do a likelihood-ratio test and directly compare my models' with the null model in terms of explained variance using Snijders and Bosker (2012)'s formula? R2 = 1 - (("model-n σ2" + "model-n τ2") / ("null model σ2" + "null model τ2)) Where : σ2 = level-two random error variance τ2 = level-one random error variance I am really sorry to bother you, and again I want to thank you very much for your help. I am a phd student and unfortunately I am not getting much support in my faculty, statistics-wise.

@junhomoon9492 3 жыл бұрын

thank you for your video! I am already subscribe to your channel. I am wondering that how to use cross-classified multilevel model. What is the difference between cross-classified and normal multilevel model? Can i find manuel and tutorial about cross-classified multilevel model?

@sugara5356 2 жыл бұрын

Thank you for your work! But I have to say the google drive links is failed (error 404). I don't know if that is my problem.

@mikecrowson2462 Жыл бұрын

Hi there, the links should be fixed now. I cannot understand how links can randomly go bad! Thanks for your patience!

@isharamawat3809 2 ай бұрын

Hi Mike I have a doubt. Which model can we run if, for example we are studying only one school? In that case, we have school level characteristics as well as student level variables.

@mikecrowson2462 2 ай бұрын

If you are only studying students at a single school you can't do multilevel analysis. You can only do a standard regression. To use multilevel, you would need data on students from multiple schools

@isharamawat3809 2 ай бұрын

@@mikecrowson2462 Thank you for the reply. So, can we take school-level variables and student-level variables in the same model? Because school level characteristics will be the same for all students. Which model will be preferred in such case?