Normal Range Calculation: Algorithm of the Probability Chart Method for the Computer

概要

論文の詳細を見る
A calculation algorithm for the probability chart method of normal range estimation is proposed. In order to get a similar normal range to the result of the eye fit regression line on the probability chart, the regression line is calculated with weighted least square method between upper limit of the class mark and the probability scale of the proportional cumulative sum using the frequency table. Normal ranges are calculated before and after logarithmic variable transformation, and they correspond to the normal ranges of normal distribution and of log-normal distribution respectively. Correlation coefficients (r) calculated with the weighted regression method can be used for selection of well fit distribution type. The result which shows larger r is selected as the well fit distribution. The normal range thus obtained is coincident well with that calculated with iterative truncation parametric method.
Japan Society of Clinical Chemistryの論文