question about small number group unit. NEED HELP! THANKS!

Welcome to the forum for MLwiN users. Feel free to post your question about MLwiN software here. The Centre for Multilevel Modelling take no responsibility for the accuracy of these posts, we are unable to monitor them closely. Do go ahead and post your question and thank you in advance if you find the time to post any answers!

Remember to check out our extensive software FAQs which may answer your question: http://www.bristol.ac.uk/cmm/software/s ... port-faqs/
Post Reply
KK123456
Posts: 4
Joined: Tue Feb 28, 2017 11:21 am

question about small number group unit. NEED HELP! THANKS!

Post by KK123456 »

Hi everyone, I have questions on 2-level multi-model with small number level-2 unit.
I am trying to study educational issue in two different areas A and B. In area A, I collected data from 20 schools with 229 individuals, and in area B, there is only data from 17 schools with 119 invididuals. I am constructing two multi-model for each area. Considering the limitation of data, I just built random slope model without adding cross level variances, and beyesian method is used.
However, the model for area A can run smoothly, and problems occurs in the model for area B. for some (not all) independent variables, when I want to see if the slope changes for different schools, it doesn't converge.

My questions:
Is there anyone could explain the above non-converge issue? Is it because the small number of schools in model? if so, why model for area A can run (only 3 schools more);

Besides, I found that even though I didn't add any level-two independents, why the model shows that variance at school-level still drop as well?

Many thanks for any help!!! 8-)
billb
Posts: 157
Joined: Fri May 21, 2010 1:21 pm

Re: question about small number group unit. NEED HELP! THANKS!

Post by billb »

Hi,
You should always build up your model step by step starting with random intercepts and pressing more when you go to random slopes instead of start to help potential convergence. Convergence is more of an issue if you smaller numbers of level 2 units and particularly if there is little variation between them. It's not simply a case of 20 vs 17 if there were no school effects in the 20 but effects in the 17 you might observe the opposite convergence issues.
Note independent variables can explain variation at all levels above the level they exist so it's perfectly OK for gender say to explain school level variance if schools with more girls do better for example.
Hope that helps,
Bill.
Post Reply