It is desirable for each clinical laboratory to have its own normal ranges. The easy
way to obtain them is to utilize the routine laboratory test values. For this purpose
many statistical methods have been proposed.
In this paper are introduced the three methods, that is, Hoffmann, Neumann and
Bhattacharya methods, and a method by the present author based on the maximum
likelihood method for estimation of parameters of truncated normal distribution.
The author described the statistical principles and the procedures in application about
the four methods, and then, applying these methods, estimated normal ranges from the
laboratory data on several test items furnished by the clinical laboratory of Busan
University Hospital and Busan Gospel Hospital. The 95% normal ranges established by
the evaluation of the results by the four methods are as follows: Serum chloride,
91-107mEq/L; serum creatinine, 0.5-1.0mg/dl; serum urea nitrogen, 6-20mg/d1; serum
thymol turbidity, 1-7 MU; serum alkaline phosphatase, 3-10 unit (Kind-King) : serum
total protein, 6-8g/dl; serum albumin, 3-5gdl ; serum globulin, 1.6-3.8g/dl; blood glucose,
65-105mg/d1.
In the precess of evaluation of the results it was found that in decreasing order of
efficacy are Bhattacharya, the maximum liklihood estimate, Neumann, and Hoffmann
method as expected. It is of the author's opinion that a better estimate of normal range
is obtained by using Bhattacharya method together with the maximum likelihood esti-
mate method or Neumann method than by anyone of them, and that Hoffmann method
does not, in general, yield good results except in some special cases.
|