.png){kind=icon}
Return
Sequential Accuracy in Parameter Estimation for Population Correlation Coefficients
.png){kind=icon}
PRE
AI
DOI：10.1037/met0000203.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Correlation coefficients are effect size measures that are widely used in psychology and related disciplines for quantifying the degree of relationship of two variables, where different correlation coefficients are used to describe different types of relationships for different types of data. We develop methods for constructing a sufficiently narrow confidence interval for 3 different population correlation coefficients with a specified upper bound on the confidence interval width (e.g.,.10 units) at a specified level of confidence ( e.g., 95%). In particular, we develop methods for Pearson's r, Kendall's tau, and Spearman's rho. Our methods solve an important problem because existing methods of study design for correlation coefficients generally require the use of supposed but typically unknowable population values as input parameters. We develop sequential estimation procedures and prove their desirable properties in order to obtain sufficiently narrow confidence interval for population correlation coefficients without using supposed values of population parameters, doing so in a distribution-free environment. In sequential estimation procedures, supposed values of population parameters for purposes of sample size planning are not needed, but instead stopping rules are developed and once satisfied, they provide a rule-based stop to the sampling of additional units. In particular, data in sequential estimation procedures are collected in stages, whereby at each stage the estimated population values are updated and the stopping rule evaluated. Correspondingly, the final sample size required to obtain a sufficiently narrow confidence interval is not known a priori, but is based on the outcome of the study. Additionally, we extend our methods to the squared multiple correlation coefficient under the assumption of multivariate normality. We demonstrate the effectiveness of our sequential procedure using a Monte Carlo simulation study. We provide freely available R code to implement the methods in the MBESS package.
Keywords:
correlation coefficient
sample size planning
accuracy in parameter estimation (AIPE)
sequential analysis
research design
Journal
IF:
7.8
Papers:
1.3K
Citations:
2.1W
Organization
Cited Papers
No cited papers available