ANOVA Designs With Subjects and Stimuli as Random Effects: Applications to Prototype Effects on Recognition Memory

To establish the generality of experimental effects, analysis of variance ( ANOVA ) designs should often treat both subjects and stimuli as random effects, but researchers seem to avoid such designs. By replicating Cantor and Mischel's (1977) study of person prototype effects on recognition mem...

Full description

Saved in:
Bibliographic Details
Published in:Journal of personality and social psychology 1987-09, Vol.53 (3), p.470-480
Main Authors: Richter, Martin L, Seay, Mary B
Format: Article
Language:English
Subjects:
Analysis of Variance
Biological and medical sciences
Experimental Replication
Fundamental and applied biological sciences. Psychology
Human
Learning
Memory
Psychology. Psychoanalysis. Psychiatry
Psychology. Psychophysiology
Psychometrics. Statistics. Methodology
Recognition (Learning)
Social research
Statistical Significance
Statistics. Mathematics
Citations: Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:To establish the generality of experimental effects, analysis of variance ( ANOVA ) designs should often treat both subjects and stimuli as random effects, but researchers seem to avoid such designs. By replicating Cantor and Mischel's (1977) study of person prototype effects on recognition memory, we demonstrate how apparently significant effects can become nonsignificant when more appropriate tests, in which recognition words are assumed to represent a random effect, are calculated. Nevertheless, uncritical use of analyses assuming random effects can also be misleading, often erring by being too conservative. We discuss the controversy over the consequences of assuming stimuli represent a random or a fixed effect and of using quasi F ratios. We recommend that researchers increase the number of stimuli to increase the generalizability of results and power and to explore substantive stimulus effects themselves. To facilitate use of ANOVA designs that incorporate two random effects, we discuss several design options and some advantages and disadvantages of each.
ISSN:0022-3514
1939-1315
DOI:10.1037/0022-3514.53.3.470