样本设计课件.pptVIP

估计比例时样本量的确定 对于简单随机抽样： 1. 确定允许误差D， d = ±.05 2. 确定置信水平CL， CL = 95% 3. 确定与置信水平相应的z值， z = 1.96 4. 估计总体的发生率P， P = 0.64? 5. 计算所需样本量n， n = P(1-P) z2/d2 = 355 6. 当样本量超过总体的5%时， nc = nN/(N+n-1) * ?????????????????????????????????????????????????????????????????????????????????????????? ??????  ?????? ?????? ?????? ?????? ?????? ?????? * 五、不回答问题 由于拒答、度量误差、数据处理误差等因素，常常产生非抽样误差。在研究中除了抽样误差以外，还应注意尽量减少非抽样误差。 * ?????????????????????????????????????????????????????????????????????????????????????????? ??????  ?????? ?????? ?????? ?????? ?????? ?????? * ?????????????????????????????????????????????????????????????????????????????????????????? ??????  ?????? ?????? ?????? ?????? ?????? ?? * 课堂讨论 1. 如果联想电脑公司委托你开展一项笔记本电脑顾客满意度调查，请确定： a. 目标总体和抽样总体？ b. 抽样单位和抽样框架？ d. 抽样方法及注意事项？ 2. 假如估计满意的顾客比率可能为70%，回答率大约为70%，要求有95%的把握保证满意率的误差不超过5个百分点， 请计算需要抽取的样本量。 * 小结 抽样的基本步骤 总体?抽样框架?抽样方法?样本量?抽样  非概率样本设计 偶遇样本；主观样本；定额样本；滚雪球样本  概率样本设计 简单随机；系统；分层；整群 样本量的的确定 经验法；统计法 回答率问题 趋势分析 * Sampling (review) Two major concerns 1. Cost Sample size Method of data collection (phone, mail, personal interview) 2. Information accuracy Sample representativeness Respondent error Administrative error Basic questions What kind of sample? How large should it be? * Sampling Problem China Light considered offering customer consulting service in which the company would help to make homes more energy efficient. To decide how much customers would like this service, they mailed questionnaires to all of their 500,000 home customers together with February bill. About 25,000 households responded. The returned responses indicated a very high level of interest in the service. So China Light hired many service workers for heavy demand. However, actual demand turned out to be low. What happened? * 参考文献 1. 麦克丹尼尔、盖兹著，范秀成等译：《当代市场调》, 原书第4版, 第12-13章，2000，机械工业出版社。 2. Malhotra, N. K. 1996. Marketing Research: An Applied Orientation, Chapters 11－12, pp.358-414. Upper Saddle