TITLE:
On Sample Size Determination When Comparing Two Independent Spearman or Kendall Coefficients
AUTHORS:
Justine O. May, Stephen W. Looney
KEYWORDS:
Fisher z-Transform, Hypothesis Testing, Power, Significance Level
JOURNAL NAME:
Open Journal of Statistics,
Vol.12 No.2,
April
27,
2022
ABSTRACT: 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.