Abstract
In order to detect assay abnormality or to analyze patients' data, assayed data are transformed using biased log transformation log (x+A), and the transformed data are observed on the two variable correlogram. One half of the normal range of each test is used as the bias A. The ratio of highly correlated tests are commonly used to detect abnormal assay results or specified pathological condition. By the observation of the correlogrom, it was found that some of the combination of two tests are not proportional each other, but showed approximately powered function relationship. Upper and lower limits of data distribution can be drawn on the correlogram and the data out of limits are analyzed. Many of them represents special pathological state of the patients, but not assay abnormality.
The data distribution model of log-normal distribution with assay error is calculated with numerical integral, and it is proved that the distribution is well simulated with biased log-normal distribution. The value of the bias (A) is proportional ro square of the assay error.
