Why there is only a single model coefficient in mixed models, if it's conditional to random effects?

by Katikarnata   Last Updated May 28, 2020 16:19 PM

I have a mixed model, where client ID is included as a random effect. It is a random intercept and slope model. I was told that mixed models have different interpretation than marginal models, and the outcomes are conditional to the random effects. I have 50 clients and test how their satisfaction changes as a function of various covariates. Say, the branch and location. From the model I get "betas" (coefficients) for 3 branches: br1 - br3. How does the interpretation of the satisfaction and the coefficient, for example, "br1" is conditional to the client, if all clients are different persons? Do we average it? This would make just one coefficient for all clients, but I read this is not averaged at all. Could anybody illustrate this with any example, how it is possible to not average over clients and still have the information "conditional to all clients" at the same time as a single number? How should it be understood, if I have a result of marginal and conditional model, both with the same number of coefficients (b0...b5), and both sets of them differ?



Related Questions


Varying slope and intercept model

Updated August 19, 2017 12:19 PM

Interpreting fixed effects coefficients

Updated August 24, 2017 14:19 PM

What are the uses of panel data regression?

Updated March 03, 2019 00:19 AM