Sosnowski, T. Analiza kontrastów: między eksploracją a testowaniem hipotez, 367-378


Abstract. The article discusses methods of comparison among means in experimental designs, where independent variable has more than two levels. These differences can be analyzed in two ways: as planned (a priori) contrasts or unplanned (post hoc, post mortem) contrasts. The choice of the method of analysis depends on the research prob-lem. If the aim is to check whether differences among means are consistent with predictions, planned contrasts will be the proper choice. In such a case the omnibus F test does not provide any useful information on the tested hypotheses and can be omitted. Instead of ANOVA, one can use multiple regression analysis, which allows for direct assessment of previously defamed contrasts. If, however, no precise predictions concerning differences among means exist, the proper method of analysis will be the omnibus F test followed (if F will turn out significant) by post hoc comparisons among means. The difference between the two kinds of contrasts corresponds to two fundamental research strategies: testing hypotheses and exploration.
Autor: Ewelina Soszyńska
Ostatnia aktualizacja: 20.04.2007, godz. 11:16 - Ewelina Soszyńska