- 作者: 姚景星
- 作者服務機構: 國立台灣大學
- 中文摘要: 我們在逐次抽樣m回中母體是隨時間而變動,第一期由母體π1 抽樣並且作估計,於第二期抽樣時母體已變到π3,我們須要由π2抽樣並且作估計,如此繼續抽樣m回。例如勞動力調查一年中在1月,4月,8月及12月共作4回抽樣及估計勞動力。因此我們須要考慮如何逐次抽樣m回才可使總損失為最小且於各回之損失亦最小。 在文中第一節為導引,於第二節作出各第i回抽樣之統計賽局且作對m回逐次抽樣之統計賽局,於第3節考慮貝氏(Bayes)及最大中最小(minimax)之策略且於第4節考慮樣本大小之選擇法 , 最後一節考是考慮完全族(Complete class)及最小完全族(mininal complete class)。
- 英文摘要: In this paper, the author has taken into consideration the problem of mtime sampling from a population which varied at each time. In the practice,we always consider the following sequential sampling; in the first time,sample is drawn from a populationπ1 and makes an estimation. After-war, in the second time sampling, population π1 is changed to π2 and wedares sampes from π2 independently to πl and make an estimation and so on.For example, we wish to estimate, sequentially the total number of employedpersons or wages in the country in January, April, August and December ofone year. So, we must consider continuously sequential sampling m timesfor which total loss is minimum and loss of each time is also minimum. In this paper, we consider m times by simple random sampling froma population which varied at each time. At δ2, we Formulate a statisticalgame in each i-th time sampling, i=1, 2,……m and a m stage game for ourconsideratian. At }3, we consider a Bayes and minimax decision and at δ4,we consider for choosing sample size. The last section, we, consider thecomplete class and minimal complete class.
- 中文關鍵字: --
- 英文關鍵字: --