CMStatistics 2022: Start Registration
View Submission - CMStatistics
B1522
Title: Accurate estimation of effect sizes: A sequential approach for scientific advancement Authors:  Ken Kelley - University of Notre Dame (United States) [presenting]
Abstract: Sequential estimation (SE) is a well-recognized approach to inference in statistical theory. In SE the sample size to use is not specified at the start of the study, and instead, the data itself and the goals of the researcher guide when a predefined stopping rule is met. Thus, rather than a fixed sample size approach to study design, which is usually based on supposed values, the final sample size in SE is unknown. This is positive because sampling stops once the goal is met, but it is negative because the necessary sample size might be larger than a researcher is able or willing to obtain. SE for accurate estimation is discussed. Then, a general effect size is discussed. Then, the two are combined into a method for obtaining an accurate estimate of this general effect size. Accurate estimation is operationalized as a sufficiently narrow confidence interval, where the goals of the research determine the desired narrowness of the confidence interval since narrower intervals illustrate less uncertainly in the estimated effect, holding the confidence level constant. Termed sequential accurately in parameter estimation, which does not require the pre-specification of supposed population parameters, as is generally necessary for power analysis in a null hypothesis significance testing framework. The premise is that if an effect size is of interest for a research study, the study should be such that an accurate effect size is obtained.