Point Estimates and Confidence Intervals in Measures of Association

Six statistics for estimating the population correlation ratio, ρ 2 , were examined for bias. The expected value of an adjusted version of the sample correlation ratio, , was found to be slightly less biased but also less consistent than ω 2 or 2 with small population effects and sample sizes. A sim...

Full description

Saved in:
Bibliographic Details
Published in:Psychological bulletin 1985-07, Vol.98 (1), p.160-165
Main Author: Fowler, Robert L
Format: Article
Language:English
Subjects:
Biological and medical sciences
Confidence Limits (Statistics)
Fundamental and applied biological sciences. Psychology
Human
Population (Statistics)
Psychology. Psychoanalysis. Psychiatry
Psychology. Psychophysiology
Psychometrics. Statistics. Methodology
Sampling (Experimental)
Statistical Estimation
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:Six statistics for estimating the population correlation ratio, ρ 2 , were examined for bias. The expected value of an adjusted version of the sample correlation ratio, , was found to be slightly less biased but also less consistent than ω 2 or 2 with small population effects and sample sizes. A simplified method for generating approximate confidence intervals for ρ 2 was developed and found to be quite efficient relative to computation time. It was suggested that the determination of the median effect size and its confidence interval directly from the observed F ratio can aid in the evaluation of research contributions and is seen as particularly useful to editors, reviewers, and meta-analysts.
ISSN:0033-2909
1939-1455
DOI:10.1037/0033-2909.98.1.160