Professor Wu Xizhi, a leadingstatistician from Renmin University of China, was invited to deliver a lectureentitled “Some misleading ideas in statistical thinking and application” on theafternoon of Nov. 28th, 2017. The lecture was chaired by Professor Yang Jing,vice director of the Center.
At the beginning of the lecture,Professor Wu Xizhi elucidated his understanding of appropriate statisticalthinking by raising several introductory questions, e.g., “Is there so-called‘right’ or ‘wrong’ concerning problems in nature?”, “Is it correct to justfollow other people’s way of doing something?”. Then he explicated theabilities and knowledge that are needed to do statistical science.
In the next part, Professor Wu expoundedon several misleading ideas in the current textbooks of statistics.
1. p< .05 as the criterion of significant difference. The appropriateness ofsetting .05 as the criterion of significant difference should depend on theresearch question itself, rather than taking it as a universal standard.
2. Themisuse of hypothesis testing. There is a logical problem in the statement “Weshould accept the null hypothesis if it cannot be rejected”. The correctversion should be “there is no sufficient evidence to reject null hypothesis”and no more.
3. Itis irresponsible to treat 30 as a criterion of large sample size. The judgmentof sample size should depend on the nature of the research.
4. Themisuse of Linear Least Squares Regression and its inappropriateinterpretations. In many cases, Linear Least Squares Regression is adopted evenwhen its mathematical assumptions (such as linearity of the model, normaldistribution of the sample data) are not fully realized.
R2 for model examination. Cross-examination is abetter way to judge whether a model is good or not, i.e., using one set of datato build the model and use another set to check its fitness.
In the final part, Professor Wu gave abrief introduction of Linear Mixed Models with Random Effects and Bayesianstatistics. The lecture was successfully ended with a Q & A session.