実験計画のためのエアロゾル計数に関する統計的考察

抄録

Optimum sample sizes for designing particle count measurements were investigated. NaCl monodisperse aerosols were counted using a condensation nucleus counter (CNC). Statistical analysis of the data from the particle count with CNC agreed with the Poisson distribution. Furthermore, a χ²-test showed that a normal (Gaussian) distribution could be applied to the data when the mean of the particle count (m) is greater than 10. This result indicated that statistical analysis for the data could be simplified by using the normal distribution. In samples with m > 10⁴, the experimental error affected the fluctuation of the data with exceeding the statistical error. Optimum sample size was influenced by m as well as the fluctuation level of the data. These results suggested that the optimum experimental design consisted of 200 data with a mean in the range 10 ≤ m ≤ 10³.