Resumen de Faut-il préférer l'analyse de variance à l'analyse de régression dans les expériences utilisant des VI continues?

Alain Méot, Patrick Bonin

• In this paper, we discuss the two possible statistical approaches when IV are continuous : analysis of variance and multiple linear regression. The statistical model common to these two approaches, and analysis of covariance is first briefly reminded. Using general arguments, followed by a psycholinguistic illustration, we discuss the advantages and drawbacks of using an experimental control of confoundings and random variability, which leads to analysis of variance, or a statistical control, which leads to multiple regression analysis. The problems to avoid for insuring the validity of the chosen approach are also underlined.