On Sample Size Determination When Comparing Two Independent Spearman or Kendall Coefficients  

作　　者：Justine O. May Stephen W. Looney Justine O. May;Stephen W. Looney(Health System Information Technology, Augusta University, Augusta, USA;Department of Population Health Sciences, Augusta University, Augusta, USA)

机构地区：[1]Health System Information Technology, Augusta University, Augusta, USA [2]Department of Population Health Sciences, Augusta University, Augusta, USA

出　　处：《Open Journal of Statistics》2022年第2期291-302,共12页统计学期刊（英文）

摘　　要：One of the most commonly used statistical methods is bivariate correlation analysis. However, it is usually the case that little or no attention is given to power and sample size considerations when planning a study in which correlation will be the primary analysis. In fact, when we reviewed studies published in clinical research journals in 2014, we found that none of the 111 articles that presented results of correlation analyses included a sample size justification. It is sometimes of interest to compare two correlation coefficients between independent groups. For example, one may wish to compare diabetics and non-diabetics in terms of the correlation of systolic blood pressure with age. Tools for performing power and sample size calculations for the comparison of two independent Pearson correlation coefficients are widely available;however, we were unable to identify any easily accessible tools for power and sample size calculations when comparing two independent Spearman rank correlation coefficients or two independent Kendall coefficients of concordance. In this article, we provide formulas and charts that can be used to calculate the sample size that is needed when testing the hypothesis that two independent Spearman or Kendall coefficients are equal.One of the most commonly used statistical methods is bivariate correlation analysis. However, it is usually the case that little or no attention is given to power and sample size considerations when planning a study in which correlation will be the primary analysis. In fact, when we reviewed studies published in clinical research journals in 2014, we found that none of the 111 articles that presented results of correlation analyses included a sample size justification. It is sometimes of interest to compare two correlation coefficients between independent groups. For example, one may wish to compare diabetics and non-diabetics in terms of the correlation of systolic blood pressure with age. Tools for performing power and sample size calculations for the comparison of two independent Pearson correlation coefficients are widely available;however, we were unable to identify any easily accessible tools for power and sample size calculations when comparing two independent Spearman rank correlation coefficients or two independent Kendall coefficients of concordance. In this article, we provide formulas and charts that can be used to calculate the sample size that is needed when testing the hypothesis that two independent Spearman or Kendall coefficients are equal.

关 键 词：Fisher z-Transform Hypothesis Testing POWER Significance Level 

分 类 号：O17[理学—数学]