| Sumario: | Three problems associated with the application of stepwise methods, which are often used in psychological and educational research to pick useful subsets of variables and to evaluate the order of importance of variables, are presented. First, commonly used statistical packages incorrectly compute the degrees of freedom in stepwise analysis, resulting in artifactually greater likelihood of obtaining spurious statistical significance. Second, stepwise methods do not correctly identify the best variable set of a given size. Finally, stepwise methods tend to capitalize on sampling error and therefore tend to give rise to results that are not replicable.
|
|---|