Co-Investigator(Kenkyū-buntansha) |
DEGUCHI Nobuhiro Keio Univ. Sch.of Med.Dept of Urology, Assistant Prof., 医学部(泌尿器科学), 講師 (90118977)
MATSUO Nobutake Keio Univ. Sch.of Med.Dept of Pediatrics, Associate Prof., 医学部(小児科学), 助教授 (50173802)
中島 淳 慶應義塾大学, 医学部(泌尿器科学), 助手 (10167546)
|
Budget Amount *help |
¥1,500,000 (Direct Cost: ¥1,500,000)
Fiscal Year 1989: ¥700,000 (Direct Cost: ¥700,000)
Fiscal Year 1988: ¥800,000 (Direct Cost: ¥800,000)
|
Research Abstract |
This study was undertaken to develop a new technique of data analysis into the field of clinical laboratory tests. Several years years ago, we noticed a possibility of analyzing feedback-network in the body utilizing Akaike's method of autoregressive modeling analysis. Encouraged by a series of favorable results obtained thus far, in the present study we tried to introduce this approach into the field of clinical laboratory data analysis. The principle is that one computes autoregressive coefficients using a series of clinical laboratory data derived from a certain patient or a healthy subject. Using the obtained autoregressive coefficients, which are proper to the subject under study, one computes Akaike's relative power contribution and builds a state equation specific to the subject. We confirmed that the relative power contribution is useful for estimating the degree of contribution of multiple variables to the fluctuation of a certain particular variable. Also, by using the state
… More
equation mentioned above, we simulated impulse response function, which enabled us to predict the mode of a certain variable when other variables fluctuate. Applying these two ways, relative power contribution and impulse response function, on endocrinologic, immunologic, and metabolic data, we proved that the present approach correctly identify the network structure of the corresponding system under study. Based on these results, we tried to establish a standard statistical method for analyzing time series of clinical laboratory data. We modified the original computer soft ware, TIMSAC, possessed by the Institute of Statistical Mathematics, so that it can be easily handled by general clinicians and laboratory technicians. Recently, we successfully transplanted the software into the personal computer system, although there remains some works for us to make man-machine interfaces smoother. If this work is completed, we will be able to distribute the software, charge free, to the laboratories which want to use it for academic purposes. Less
|