| Summary: | 碩士 === 國立陽明大學 === 公共衛生研究所 === 94 === In analyzing data, the Pearson correlation coefficient is used for measuring the degree of linear association between two continuous variables, X and Y. When both two variables are affected from an extra variable called extraneous (confounding) variable, Z, the Pearson correlation result will be biased.
Therefore, for controlling for extraneous effect, the partial correlation coefficient is often used. And, it is realized that only the linear association part is removed while patial correlation applied. In this study, combinig “stratification” and “(polynomial) regression model” techniques, three corrected methods are proposed.. The formula of Perason correlation coefficient is corrected from ”group mean”、”fitted value of group regression”、”fitted value of polynomial regression”.
In this study, the efficiency of proposed methods are investigated through data simulation. The results reveal that the “fitted value of group regression” is the best efficiency method. Applying the proposed methods in fetal data, the age-corrected Pearson correlation coefficient between “femur length” and “predicted weights” is about 0.5, which is very different from the uncorrected value, 0.9.
Meanwhile, when the extraneous effect is along with heterogeneous variance problem, a corrected technique also is proposed in this study.
|
|---|
