統計技法을 活用한 辨證定量化 硏究

다국어 초록 (Multilingual Abstract)

For the work of standardization of Bian Zheng [辨證], it is necessary to quantify both patients’ symptoms and signs. If the quantity of data acquired from patients is too big for us to apply a simple statistical method, a specific statistical analytic method is needed for correct interpretation of those data.
Principal components analysis(PCA) is typically used in grouping variants when the variants used in specific studies are too much diverse. In the standardization of Bian Zheng [辨證], the statistical methods which have been used have difficulty in dealing with the variety of patients’ signs and symptoms. In this study, we applied principal components analysis to improve the efficiency of Bian Zheng [辨證]’s standardization techniques. With principal components analysis, we also applied correlation analysis and regression analysis method.
We made a questionnaire consisted of symptoms representative of Han Zheng [寒證], Re Zheng [熱證]and Xu Zheng [虛證]from 162 subjects. The number of questionnaire items for symptoms were numbered twenty four, and they were answered in the form of Likert-like five-points score. With questionnaire, we performed Pulse rate variability tests including total power, very low frequency, low frequency, high frequency, and we also performed autonomic function tests. Afterwards, data items for symptoms expressed in Likert-like five-points score, log-transformed Pulse rate variability indices and autonomic score were analyzed by correlation analysis, principal components analysis, and regression analysis respectively.
In correlation analysis, mean heart rate, ln(TP), ln(VLF), ln(LF) and ln(HF) were negatively correlated with age. In intercorrelation of time domain indices, SDANN was negatively correlated with mean heart rate. In intercorrelation of frequency domain indices, SDANN was negatively correlated with mean heart rate. In intercorrelation of frequency domain indices, ln(TP), ln(VLF), ln(LF) and ln(HF) were positively correlated with each other. ln(TP) and ln(HF) were negatively correlated with ln(LF/HF ratio). Autonomic score was negatively correlated with ln(TP), ln(VLF) and ln(LF). In correlation analysis of Pulse rate variability indices with twenty four data items for symptoms, ln(TP), ln(VLF) and ln(LF) were negatively correlated with Han Zheng [寒證] item and Xu Zheng [虛證] items. R-R interval was negatively correlated with Re Zheng [熱證] items.
Before exercising the factor analysis of items for symptoms, we performed correlation analysis to see if each item for symptoms was fit to be inserted to principal components analysis. As a result, the correlation coefficients of five data items for symptoms were all below 0.3 and these data items were excluded in principal components analysis. In principal components analysis of nineteen data items for symptoms, Han Zheng [寒證] and Re Zheng [熱證] were separated in the form of two factors respectively. Xu Zheng [虛證] maintained a single factor. Afterwards, data items for symptoms, Pulse rate variability indices and autonomic score were factor-analyzed by principal components analysis. Before exercising principal components analysis of data items for symptoms including Pulse rate variability indices and autonomic score, we performed correlation analysis to see if Pulse rate variability indices and autonomic score were fit to be inserted to principal components analysis. As a result, the correlation coefficients of Pulse rate variability indices and autonomic score were both below 0.30, and additional statistical analyses were not applied to them.
Finally, we performed regression analysis on data items for symptoms, Pulse rate variability indices and autonomic score. We extracted all regression equations of ln(TP) and ln(VLF) which included Xu Zheng [虛證] factor as an independent factor.