Using Quasi F to prevent alpha inflation due to stimulus variation

    loading  Checking for direct PDF access through Ovid


Suggests that the nominal alpha level may be very inflated in much of the published literature where the conventional F test is used. This alpha inflation is often caused by ignoring stimulus variation or treating it as a fixed effect. This problem is illustrated in a variety of areas, and the use of Quasi F ratios as a means of achieving generality over both Ss and stimuli is discussed. Monte Carlo experiments are reported that examine the performance of the Quasi F in a variety of realistic situations in which the data violate distribution and homogeneity of variance assumptions. In general, the Quasi F has proved to be robust, and it is suggested that more general use of such tests would greatly increase the reliability of the published data base. (13 ref) (PsycINFO Database Record (c) 2006 APA, all rights reserved)

Related Topics

    loading  Loading Related Articles