在規劃研究設計時,研究人員往往忽視樣本多寡的問題,因而造成統計推論上的嚴重偏差。忽視樣本的結果,將使得:(1)由於樣本數不足,雖已達到臨床上的顯著差異,卻冉法達到統計上的顯著差異;(2)因樣本過多,雖研究結果沒有臨床上的意義,卻達到統計上的顯著差異及(3)由於樣本有限,雖然研究結果皆達臨床及統計上的意義,但檢力過低無法肯定其對立假說是否成立。本文提出了在單一及雙常態母體平均值檢定、二項分佈檢定、邏輯斯迴歸分析、變方分析世代死亡率之樣本計算公式,並援引數例說明之,俾提醒研究人員能藉此瞭解樣本數之重要性並妥爲運用之。
Researchers always ignore the problem of sample size when they are planning a research design. This might lead to the incorrect result of statistical inference. In this paper, we present three examples to describe the problems when sample size are small or too large in research design. Also, we provide the formulas of sample size and some examples for comparing the means of one and two normal distribution, binomial test, logistic regression, ANOVA and mortality of chort study to remind researchers paying attention to the importance of sample size for research design. Although there are limitations on the formula of sample size related to some statistical analysis, this paper do provide a guideline for handling the quality of research design by carefully considering the requirement of sample size.