Would you explain a practical situation where random effect model is more appropriate than the fixed effect model?
More information would help us discern the structure of your data. Yet generally speaking, if your data are clustered (e.g., correlated or nested), you should use mixed-effects models that contain both fixed and random effects. For example, you obtained outcome variables:
- repeatedly from the same set of subjects. In this case, observations are nested within subjects and are correlated; or
- from students nested within classrooms or patients nested within hospitals. Then the observations taken from the students/patients within each classroom/hospital can be correlated.
Ignoring such a clustering structure of your data can bias your estimates and/or inflate type-I error rates (e.g., Moerbeek, 2004; but see Van Landeghem et al., 2005).
If your data are obtained from subjects only once, and they are not nested within any grouping/structural variables, you can use fixed-effect models.
- Moerbeek, M. (2004). The consequence of ignoring a level of nesting in multilevel analysis. Multivariate Behavioral Research, 39, 129–149.
- Van Landeghem, G., De Fraine, B., & Van Damme, J. (2005). The Consequence of Ignoring a Level of Nesting in Multilevel Analysis: A Comment. Multivariate Behavioral Research, 40, 423-434