ABSTRACT
The usual method to estimate the parameters of the multilevel regression model is
Maximum Likelihood estimation. This produces parameter estimates and asymptotic
standard errors, which can be used to test the significance of specific parameters, or to set
a confidence interval around a specific parameter. Chapter Three mentions alternatives to
this standard approach to estimation and testing. This chapter discusses some several
alternatives in more detail: the profile likelihood method, robust standard errors,
bootstrapping, and Bayesian methods.
