If you have data where the observations are not independent due to nesting or clustering, you may need a multilevel model. Another scenario that would require a multilevel model is if you have data where observations have been gathered multiple times on the same subject (a.k.a., longitudinal data or repeated measures).
Here are a few examples:
- Student math scores, where students are nested within classrooms and then further nested within schools;
- Automobile loan default (yes/no), where loans are nested within automobile dealerships, which are nested within relationship managers;
- Measures of reforestation (such as number of species growing), taken annually for several years at specific locations after a forest fire;
- Counts of defected silicon wafers, where the wafers are nested within manufacturing site; and
- Patient outcomes where patients are nested within doctors and doctors are nested within clinics.
Whenever observations are clustered, they will be correlated. If your model doesn't account for these correlations, your inference may be compromised. Another advantage of multilevel models is that, in addition to accounting for the correlations during modeling, you can obtain estimates of the degree of correlation among your observations.
You can learn more about multilevel models by signing up for our course. If you’re attending Analytics 2013 in London, we’re also offering a training session on June 17 and 18.
5 Comments
Pingback: When is a Multilevel Model not appropriate? - The SAS Training Post
Hello,
Can you tell me when mlm is not appropriate? I have data that by design is clustered but the random intercept in the null model is not significant. I have seen advice that says when this is the case all of your level 2 variation and thus the model for level 2 is over-fitted. I have heard others say that that doesn't matter use mlm anyway.
Any response is most appreciated.
Great question Candace--I'll post a blog with a response shortly.
Hi Candice,
Chris created a new blog post to help answer your question: http://blogs.sas.com/content/sastraining/2013/04/24/when-is-a-multilevel-model-not-appropriate/
Let us know if you have any other questions. Thanks for reading.
Pingback: What is a multilevel model?