Sample size requirements for multiple regression interval estimation
提出一个简单、准确且通用的方法，用于近似计算获得期望精度的多元相关系数平方置信区间所需的样本量，同时给出估计非标准化回归系数所需样本量的简便方法。
Abstract Sample size planning is one of the most important issues in the design of a study. Simple and accurate sample size formulas for a desired confidence interval width have been developed for many statistical procedures, but a simple and accurate sample size formula for the squared multiple correlation has been a notable exception. Several rule‐of‐thumb sample size recommendations for a multiple regression analysis have been proposed over the years but none are satisfactory. Other approaches have focused on the construction of elaborate tables of sample size requirements, but these tables are both unwieldy and inadequate. We present a simple, accurate, and general method of approximating the sample size requirement for obtaining a squared multiple correlation confidence interval with desired precision. We also present a simple method for approximating the sample size needed to estimate unstandardized regression coefficients with desired precision. Copyright © 2010 John Wiley & Sons, Ltd.