| Sumario: | We discuss several linear and nonlinear regression models and their appropriateness in modeling a set of data from a cohort study of the use of medical services in a prepaid medical plan. Although computational considerations once forced researchers to use linear models almost exclusively, modern computer hardware and efficient algorithms have made it possible to use nonlinear models almost as easily as linear ones. Thus, in choosing a regression model, a researcher should give strong consideration to how well the model reflects (presumed) intrinsic properties of the phenomena under study, even if the consideration leads to the choice of a nonlinear model. Modern computer programs for nonlinear minimization should be employed to fit the model. The model should be judged on how well it describes the data and how reasonably it predicts the future.
|
|---|