Abstract: When using traditional classical statistic analysis to extract sensitive grain size of the East Asian monsoon, misunderstandings are often raised due to neglect of closure effect caused by component data; therefore, the correspondence of the proxy and other climate indicators is always unsatisfactory. We take core H07 which is from the central mud area of the South Yellow Sea as the object of this research, and try to use the method of centralized logarithm ratio (clr) transformation for grain size data preprocessing to eliminate closure effect. Then, the method of factor analysis is applied to extract the sensitive grain size from the transformed data. Further on, we select the factor score as a proxy for the East Asian winter monsoon strength. Combining with AMS¹⁴C dating data, East Asian winter monsoon variation over the past 6200 years is reconstructed. The variation of the East Asian monsoon can be divided into five stages: (1)6.2~4.99 kaBP, stage of violent fluctuation in high-level; (2)4.99~3.66 kaBP, stage of slight fluctuation in low-level; (3)3.66~0.87 kaBP, stage of metastable in low-level; (4) 0.87~0.14 kaBP, stage of sustained enhancement; (5) 0.14 kaBP up to now, stage of significant weakening. There is a fine correspondence between the proxy and other climate indicators, and it is proven that applying the method of clr transformation to sensitive grain size extraction is feasible. This method can effectively reveal the history of East Asian winter monsoon variation in the eastern part of China.