Conventional point estimation does not guarantee the confidence of estimated value. In this paper, the authors first define a confidence factor (absolute confidence factor
C (Δ), relative) confidence factor
c (δ) as a measure of confidence, and then show an estimating method by which the (absolute or relative) confidence factor of estimated value is kept at a specified value γ. Especially, there is shown a decision rule (criterion) which determines a necessary and sufficient minimum sample size
n. Although there are some cases where
n is known before sampling, in many cases
n cannot be found before samling, because the decision formula is the function of samle values. As an estimating method in the latter cases, the authors propose a new method called “sequential estimation”, in which, starting from a properly small size (initial sample size)
nt, observers gradually increase the sample size
n, checking at every sampling stage if the decision formula is going well, and stop sampling and estimate the parameter when the decision formula is first satisfied. Also shown is the method to decide the initial sample size
nt.
抄録全体を表示