Evaluation of a Transformation Method for Estimation of Normal Range

[摘要] We have evaluated a recently proposed method for estimating normal limits by prior transformation of data to gaussian-like form. Computer simulation techniques are used to generate random data from frequency distributions having various shapes. For each such distribution, the transformation method is compared with the method of nonparametric percentile estimation. We conclude that the nonparametric method should continue to be the method of choice unless data have a nearly symmetrical frequency distribution. The present state of methods for estimating normal ranges is discussed.

[学科分类] 过敏症与临床免疫学