Sample Size Calculation for Paired t-test Using p-value and Confidence Intervals When Standard Deviation of Difference is Not Given.

Abstract

Sample size calculation is an important step for carrying out any research. Most of the sample size estimating utilities/applications use the standard deviation of differences in their formula for paired design, which usually is not reported in the published articles. Researchers are left to use the pooled standard deviation (SDp) instead of using the standard deviation of differences. The standard deviation of differences gives a more accurate estimate of sample size for a paired design. The standard deviation of differences can be determined from the p-value or confidence intervals (CI), which are usually reported in the published articles. In this paper, we have described and illustrated with an example how to determine sample size from the p-value or CIs. Readers may follow the given example using scripts for software R.