EDUCATIONAL AND PSYCHOLOGICAL MEASUREMENT FAN AND THOMPSON CONFIDENCE INTERVALS FOR EFFECT SIZES CONFIDENCE INTERVALS ABOUT SCORE RELIABILITY COEFFICIENTS, PLEASE: AN EPM GUIDELINES EDITORIAL

Confidence intervals for reliability coefficients can be estimated in various ways. The present article illustrates a variety of these applications. This guidelines editorial also promulgates a request that EPM authors report confidence intervals for reliability esti-mates whenever they report score reliabilities and note what interval estimation methods they have used. This will reinforce reader understanding that all statistical estimates, including those for score reliability, are affected by sampling error variance. And these requirements may also facilitate understanding that tests are not impregnated with invariant reliability as a routine part of printing. During recent stewardship of this journal, the editors have promulgated a number of “guidelines editorials ” for prospective authors. These have cov-ered issues such as 1. effect sizes, including explicit requirements that effect sizes be reported along with p values (Thompson, 1994); 2. score reliability, including requests that tests not be described as reliable and that what Cohen (1994) termed “nil ” nulls (e.g., H0: α = 0 as opposed to a non–nil null such asα>.75) not be used to test reliability or some validity co-efficients (Thompson, 1994); 3. stepwise analyses, in the form of proscribed reporting of these method