A fixed effects model treats the effects associated with the observed factor levels or units as unknown constants to be estimated. In a one-factor model,
 |
the parameters 
describe the particular levels included in the study. A constraint such as 
or a reference-level parameterization makes
the representation identifiable.
The term contrasts with a random effects model, in which level effects are modeled as random variables drawn from a population and their variance is a variance component. Which interpretation is appropriate depends on the sampling and inferential goals, not only on the algebraic form of the fitted model.
