Multilevel modeling using STATA (updated 2/9/18)

This video provides an introduction to using STATA to carry out several multi-level models, where you have level 1 and level 2 predictors of a level 1 outcome variable. The video begins with a random intercept model and concludes with a model incorporating Level 1 and Level 2 predictors, along with varying intercepts and slopes. Some discussion of cross-level interaction is provided.
Data for this video can be downloaded at: drive.google.c...
The Excel calculator for computing significance tests for variance components can be downloaded here:
drive.google.c...
You can also download the notes I go over here: drive.google.c...
DEAR VIEWERS: I HAVE RECENTLY UPLOADED A MORE UP-TO-DATE (and higher quality) VIDEO ON MULTILEVEL MODELING USING STATA HERE (2020 VIDEO): • Multilevel regression ... . THAT VIDEO ALSO GOES INTO TESTING AND PROBING INTERACTION EFFECTS, AND UNDERNEATH THE VIDEO DESCRIPTION YOU WILL FIND LINKS TO THE DATA FILE AND A TEXT FILE CONTAINING COMMANDS. ONCE AGAIN, PLEASE CHECK OUT • Multilevel regression ...

[@indirakhadjieva2108]

Dear Dr.Crowson! Thank you for the invaluable insights and knowledge you shared in this video on multilevel hierarchical models. The clarity and depth of your explanation greatly enriched my understanding of this complex topic.

[@meganendres7528]

This is the best walkthrough I have seen. I went through it step-by-step with my own data to learn it - awesome!

[@mikecrowson2462]

Hi Megan, I'm so glad you found the video useful. Just so you know, I have a much more recent video here (kzbin.info/www/bejne/a6SlZWyQic13itE). Underneath the video description is a powerpoint and data for you practice with as well. Please feel free to share. Best wishes!

[@claire2247]

Thank you Mike! I will not for anyone watching though, that you don't necessarily have to use the calculator he made in excel. If the confidence interval doesn't include zero, you can be fairly certain that the estimate is significant (95/100 times the estimate will fall in that range).

[@leolele007]

Great video! To calculate p-values from estimate and standard error you can also use the following command instead of the calculator: display 2*(normal(-(Estimate/StandardError)))

[@mikecrowson2462]

Thanks for visiting and for the tip! I'll keep this in mind in the future! Cheers!

[@tanjamayer6469]

Thanks for your useful videos! I wish actually doing statistics with stata had been explained to me so vividly and clear in university.

[@mikecrowson2462]

Thanks for your comment, Tanja. Glad you are finding them useful! Good luck with your research!

[@luisortizgervasi3820]

Congratulations, Mike. It seems excellent, very well-explained and clear, in my view. Many thanks for sharing.

[@unigruppe7474]

Cheers! That is the most useful explanatory video about statistics I have ever watched!

[@nicky103131]

This is the most straight forward video for me, after watching >10 of them... Thanks a lot!

[@jenniferrous3280]

Thank you for the detailed example walking through the progression. It helped me get closer to solving my question.

[@allekoennentechnik]

Awesome video! Exactly what I needed, thank you!

[@semsdincer1464]

Hello sir, First of all, thank you very much for your contribution to the student from all over the world. Secondly, could you please recommend an introductory textbok for multilevel modelling?

[@Cuezaltzin]

Excellent lecture. Thank you.

[@mcortinatoro]

Thank you very much for this video!

[@El_Pancho_Alvarez]

Amazing video. Thank you so much. Could you delve further into graphics for the residuals for the different models and how to graph the random slopes to see graphically if they are fading inwards or outwards?

[@mikecrowson2462]

DEAR VIEWERS: I HAVE RECENTLY UPLOADED A MORE UP-TO-DATE (and higher quality) VIDEO ON MULTILEVEL MODELING USING STATA HERE (2020 VIDEO): kzbin.info/www/bejne/a6SlZWyQic13itE . THAT VIDEO ALSO GOES INTO TESTING AND PROBING INTERACTION EFFECTS, AND UNDERNEATH THE VIDEO DESCRIPTION YOU WILL FIND LINKS TO THE DATA FILE AND A TEXT FILE CONTAINING COMMANDS. ONCE AGAIN, PLEASE CHECK OUT kzbin.info/www/bejne/a6SlZWyQic13itE

[@SuvraGhosal]

can you plzz explain multinomial multilevel analysis similarly ??

[@maryamsalari3490]

Hi. Thank you a lot. I need the data set that you have used in this video. Would you please send it for me?

[@asayealamneh4085]

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

[@amanrastogi5184]

Sir, your explanation of the topic was too god, but please do something regarding the things that you write, its difficult to understand completely.

[@mikecrowson2462]

Hi there. thanks for visiting and your feedback. Just an FYI, I have a much newer video (with additional supporting materials such as data and a text file containing commands) on multilevel regression using Stata here: kzbin.info/www/bejne/a6SlZWyQic13itE . Best wishes!

[@shamsishekari2207]

Hi Mike, Thank you very much. I was wondering how you are calculating Z value and p values in your excel sheet. I couldn't find any straightforward formula for Z value. Essentially I am interested in your formulas in excel sheet. Thank you, Shamsi

[@strategicmanagement7406]

Thanks for the video. Wondering if you could post the calculator you have developed to test the significance of level variances for downloads. Thanks

[@ph.d.mayanksingh5769]

hello Mike. I was using the bayesian multilevel model and getting error (Too many MCMC estimates to save in a simulation File). Can you help to me to solve this issue. Command Is: bayes, mcmcsize(10000) burnin(1000) thinning(10): melogit marri || v024: || sdist || v001 || v002:, level(95) here marri: dependent variable coded into 0 and 1 v024: state sdist: district v001: cluster v002: individual

[@ludovico86]

I really appreciate your videos. In uni I felt there is often a disconnect between the output and the theory, which you explain really well. Any chance of you taking on mixed effects logistic regression models?

[@mikecrowson2462]

Hi. Thanks for watching and your feedback. I haven't worked up a Stata version of it, but I do have an SPSS version of this. You can go here to watch: kzbin.info/www/bejne/mIGUeYOLaNikrNk A Stata overview is on the horizon.

[@enkii82]

Hi, I have a question about weighing. Let say I have 3 schools with one has 20, another has 8 and the third has 12 students. So the number of students across the cluster is not equal. Doing the multi-level will just get an average of each cluster. How to do it on STATA to include the weight in multilevel mixed effect? Thanks.

[@tawilk]

is there a way to get prediction index values from these models? I was hoping this 2 level approach would have improved my classification (sensitivity) values.

[@barkebille1]

Thank you for this video! I´m using STATA 15 myself, and I´m trying to figure out one question about mixed effects models, perhaps you can help me? If I run a mixed command with a continous dependent variable, one fixed dichotome predictor of the dependent variable and a fixed time variable, and on the right side of the || I add the id as random intercept and time as random slope. Now, my question is: This model is constructed with both fixed effects and random effects, but as far as I have understood, the results found in the fixed part of the outcome are the numbers which matters, as they give the information about the main effects? Since the model is run with random effects included, does this mean that the fixed effects have been adjusted for the randomly varying intercepts and slopes?

[@KamarulImran]

Hi Mike, excellent video. May I ask, when you used the mixed menu, the codes for RS was 'mixed y x1 x2 || schcode: || schcode:: ses', but what pasted your codes from MS Word it was 'mixed y x1 x2 || schcode:ses'. Do they mean the same? Will the results be similar?

[@hamedkashefi9656]

How do we make a output tabel for the multimode

[@asayealamneh4085]

could you have a video about the multilevel binary logistic regression model?

[@mikecrowson2462]

Hi there. I have a new video on this topic that I've uploaded on SPSS (kzbin.info/www/bejne/qKC6hYiCntKBhs0). Although you may be a Stata user, the video and its accompanying Powerpoint (drive.google.com/open?id=16UJsWJodaVFdxJesu7OTQFgGWtrsITzv) might be useful for learning about it. I am currently working on a new presentation related to Stata (but don't have an eta right now). If you want to want to know some of the Stata commands, you can download something I worked up here: drive.google.com/open?id=1_VJMwGFGHDz4zIBbL5y8HkbPECMY9swe Here's a Stata file of the data if you want to play with the data: drive.google.com/open?id=1O9Ii8UMKdWcE4oVtGLbnc_bzGfPYQ8GQ One other thing, here's a new presentation I created on standard multilevel modeling using Stata. It includes a Powerpoint with it as well (download under the video at kzbin.info/www/bejne/a6SlZWyQic13itE) Cheers!

[@Thnawiga]

Hi I was wondering if I need to do a multilevel logistic regression will it be done the same way? if not is there a video that explains it?

[@mikecrowson2462]

Hi there. Unfortunately I don't have a video on this topic for Stata. But I do have a video for SPSS (kzbin.info/www/bejne/mIGUeYOLaNikrNk) . I believe that Stata has the ability to do this. You might check out multilevel mixed effects regression Stata manual at www.stata.com/manuals/me.pdf . Good luck!

[@gulafshan9388]

Sir, is there any standard value of ICC that is required to go for MLM. What if it is low i.e just 6 or 7%?

[@listub3231]

same question here...

[@justicelove9605]

Thank you!!!

[@mikecrowson2462]

Hi Justice, thanks for visiting. By the way, please consider checking out my newest video on multilevel modeling with Stata at: kzbin.info/www/bejne/a6SlZWyQic13itE best wishes!

[@CalzOmon]

Thank you