体育统计学新视角:数据级数推断
发布时间：2019-05-29 09:29
【摘要】：抽样研究通常是对抽取的样本进行测量,根据测出的样本数值对总体参数值进行推断,假设检验、置信区间和数据级数推断的方法都被运用来进行这一研究推断。本文采取了一个体育学的抽样研究案例,分析和探讨了以上三种方法在做出研究推断时的优缺点。主要结果包括:基于传统的假设检验提供的p值进行推断时,推断结果具有较大的误导性,只能获得一个"二元结果":显著或不显著,同时还面临着"I型误差"和"II型误差"带来的干扰。置信区间的添加可以对p值进行一定的补充,但并不能对研究结果推断提供实质性的、进一步的信息。通过加入一个"最小有意义阈值",数据级数推断方法能给出一个"三元结果":有意义正差异/相关、有意义负差异/相关和无意义微小差异/相关,并基于置信区间推算出总体值落在每一个结果区域的可能性,从而能让定量研究结果的推论更加丰富和准确,同时大大降低I型误差率和II型误差率,减少发表偏倚。
[Abstract]:Sampling research is usually to measure the extracted samples, according to the measured sample values to infer the overall parameter values, hypothesis testing, confidence interval and data series inference methods are used to carry out this research inference. This paper takes a sample study case of physical education, analyzes and discusses the advantages and disadvantages of the above three methods in making research inference. The main results include: when inferred from the p value provided by the traditional hypothesis test, the inference result is misleading and can only obtain one "binary result": significant or not significant. At the same time, it is also faced with the interference caused by "type I error" and "type II error". The addition of confidence interval can supplement the p value to a certain extent, but it can not provide substantive and further information for the inference of the research results. By adding a "minimum meaningful threshold", the data series inference method can give a "ternary result": meaningful positive difference / correlation, meaningful negative difference / correlation and meaningless small difference / correlation. Based on the confidence interval, the possibility that the overall value falls in each result region is calculated, so that the inference of quantitative research results can be more abundant and accurate, and the type I error rate and type II error rate can be greatly reduced, and the publication bias can be reduced.
【作者单位】： 华南师范大学体育科学学院;Institute
【基金】：广东省普通高校青年创新人才项目《足球大数据分析方法探讨与实证运用》,项目号:2016WQNCX015
【分类号】：G80-32
，
本文编号：2487830
[Abstract]:Sampling research is usually to measure the extracted samples, according to the measured sample values to infer the overall parameter values, hypothesis testing, confidence interval and data series inference methods are used to carry out this research inference. This paper takes a sample study case of physical education, analyzes and discusses the advantages and disadvantages of the above three methods in making research inference. The main results include: when inferred from the p value provided by the traditional hypothesis test, the inference result is misleading and can only obtain one "binary result": significant or not significant. At the same time, it is also faced with the interference caused by "type I error" and "type II error". The addition of confidence interval can supplement the p value to a certain extent, but it can not provide substantive and further information for the inference of the research results. By adding a "minimum meaningful threshold", the data series inference method can give a "ternary result": meaningful positive difference / correlation, meaningful negative difference / correlation and meaningless small difference / correlation. Based on the confidence interval, the possibility that the overall value falls in each result region is calculated, so that the inference of quantitative research results can be more abundant and accurate, and the type I error rate and type II error rate can be greatly reduced, and the publication bias can be reduced.
【作者单位】： 华南师范大学体育科学学院;Institute
【基金】：广东省普通高校青年创新人才项目《足球大数据分析方法探讨与实证运用》,项目号:2016WQNCX015
【分类号】：G80-32
，
本文编号：2487830
本文链接：https://www.wllwen.com/jiaoyulunwen/tylw/2487830.html
