Expanded analysis of Simpson’s Paradox: A Case Study

Simpson’ s paradox is a famous paradox in economic field. It explains the abnormal statistical results with respect to two different variables. That is to say, in some kinds of census, the trend of the whole group may not be able to represent the separate groups. The application of it includes enrollment of universities, success rate of treatment, the effect of smoking on life span, admissions policies and so on. The paradox plays a significant role in statistical analysis, because in some terms, it can decide the accuracy of the data sets. The paper aims to seek possible solutions to reduce the uncertainties emerged in the paradox with respect to mathematical methods. In this paper, theoretical and real case analysis are presented. Several representative cases are analyzed, and previous methods are included. The author raises an improvement for the methods above. The conclusion is that if proper weight is added, there will be little error in spite of the occurrence of Simpson’ paradox.