体育科学研究中多元统计分析常见问题及对策
Common Problems of Multivariate Statistical Analysis in Research of Sport Science and Their Solutions

摘要：为了规范体育科学研究方法，提高多元统计分析的科学性，针对统计描述、统计推断和综合评价3个体育科学研究领域中存在的典型问题，分析了问题的性质和原因，提出了解决问题的思路和具体建议。指出：在统计描述工作中，要明确描述对象，确定描述的内容，合理选择和使用统计指标，重视描述性评价；对于多元均值的推断，需要明确统计总体和个体，理清需要推断的问题，运用多元检验统计量检验多元均值向量，运用联合置信区间检验各个分量及其线性组合；关于体育综合评价，建议运用主成分或因子分析法分析评价指标集的内部结构，采用专家法对主成分或因子赋权。
Abstract: In order to regulate the scientific research methods in sport and improve the scientificality of multivariate statistical analysis, this paper points out the typical problems in three scientific research areas of sport, including statistical description, statistical inference and comprehensive evaluation. The nature and reasons of the problems are analyzed and the ideas to solve the problems and specific suggestions are put forward. In the studies of the statistical description, it is required that the object of description should be clear, the content of description should be determined, the statistical indicators should be chosen and used reasonably, the descriptive evaluation should be taken seriously. For the inference of multivariate mean vectors, the statistical population and individual must be clear, and the problems of inferring should be clarified, the multivariate test statistics should be used to test multivariate mean vectors, and the joint confidence interval should be used to test each component and its linear combination. With respect to the comprehensive evaluation in sports, the proposal is put forward that principal component analysis or factor analysis can be applied to analyze the internal structures of the evaluation index set, and the weight coefficients of each principal component or factor can be determined by experts