Sampling & Power Analysis 终极背诵版
1. 基本概念
• Population:总体
• Sample:样本
• Sampling unit:抽样单元
• Parameter(总体):N, μ, σ
• Statistic(样本):n, mean, SD
2. Probability Sampling 概率抽样
(可推论总体 generalizable)
1. Simple Random Sampling 简单随机:等概率
2. Systematic Sampling 系统抽样:随机起点 + 固定间隔
3. Stratified Sampling 分层抽样:先分层,再每层抽样
4. Cluster Sampling 整群抽样:先分群,再抽群
5. Multistage Sampling 多阶段抽样:两种以上结合
Non-probability Sampling 非概率抽样
(不可推论总体 NOT generalizable)
1. Convenience 方便抽样
2. Purposive / Judgment 判断/目的抽样
3. Quota 配额抽样
3. Sampling Bias 抽样偏差
(systematic difference → low external validity)
• Coverage bias:覆盖偏差,抽样框不全
• Non-response bias:无回应偏差
• Self-selection bias:自选择偏差
• Survivorship / Healthy User bias:幸存者偏差
• Under/over-coverage:覆盖不足/过度
4. Power Analysis 功效分析
• Power = 1 – β
• α:Type I error 假阳性
• β:Type II error 假阴性
影响功效的4个因素:
• Effect size 效应量
• Sample size 样本量
• α level
• Statistical test
效应量指标:
• r(相关), d(均值差), f/η²/R²(回归/ANOVA)
样本量经验法则(r):
• Small (r=0.20): n ≥ 300
• Medium (r=0.50): n = 60–100
• Large (r=0.80): n ≥ 30
5. Sampling Decision Steps 六步
1. Define target population
2. Select sampling frame
3. Choose sampling method
4. Determine sample size
5. Collect data
6. Evaluate response rate
6. 批判题万能句式(直接套)
• This is non-probability sampling, so findings cannot be generalized.
• There is self-selection / coverage bias, which threatens external validity.
• Sample size may be insufficient to achieve adequate statistical power.
• The sampling frame is incomplete, leading to under-coverage.
