Optimum Sample Size for Estimating Coefficients of Variation in a Normal Population

概要

論文の詳細を見る
The coefficient of variation is widely used as a measure of variation and as a sampling survey statistic. Little information is available in the literature, however, on the optimum sample size for estimating the coefficient of variation of a population with reasonable accuracy. We attempted to identify a method of normal approximation for determining the sample size for estimating the coefficient of variation of a population. Based on Haldane's method, we estimated the coefficient of variation of a sample approximately corrected for bias by a correction factor as follows: V= (1+1/4n) ⋅s/m, where n is the sample size, s is the sample standard deviation, and m is the sample mean. A standard error of the coefficient of variation (Sv) and its 95% confidence limit (2L) were calculated based on normal approximation of the coefficient of variation of a sample as follows: Sv = (V2/2n⋅ (n/ (n-1) +2V2) ⋅ (1+1/4n) 2) 0.5, L = t⋅Sv, where n is the sample size, t is the Student's t-value for 5% with a degree of freedom (n-1), and V is the coefficient of variation of a sample. The optimum sample size required for a 95% confidence level and a limit of error was computed as the minimum sample units satisfying the above formulas. We also evaluated the usefulness and reliability of the estimation through a simulation study for determining the optimum sample size for coefficients of variation. Random samples of normally distributed random variables were generated for 4 sets of sample sizes and coefficients of variation values with each 3, 000 replications, and 3, 000 random samples from data on rice anther length were evaluated with 95% confidence. These sampling experiments showed precise estimates in 2.03 - 4.54% of the samples outside the desired limits of error.
日本熱帯農業学会の論文

Optimum Sample Size for Estimating Coefficients of Variation in a Normal Population

スポンサーリンク

概要

日本熱帯農業学会 | 論文

スポンサーリンク