Abstract
The study was made on the significance of distribution type on the normal range calculation. Computer simulated random data were generated from the population after variable trans-formation (log x, 3X, X, x2, x3 and x) and they were subjected to distrlbu tion type caluclation by maximum likelyhood method (MLH) or minimum skewness method (MSK).No evident difference of the distribution typing between MLH and MSK is observed, but MSK calculation is faster than MLH.When number of data (N) is larger than 1000, the calculated distrlbution iscoincident well with that of the population, but when N is less than 500, it was different from the expected distribution. Contamination of a small amount abnormal data effects on the selection of the distribution type, and misleads to errorneous normal range calculation. The result indicates that proposal of many distribution model in the normal range calculation is inadequate if N is Iess than 1000. and only normal and log-normal distributions are enough as the model distributions.
