论文部分内容阅读
通常对于正偏态分布资料采用几何均数法处理,但对于那些经对数转换后不能变为正态分布的资料,这样做却缺乏理论依据;而采用百分位数法处理,又不能充分利用资料内部所蕴藏的信息.本文从理论上阐明用威布尔分布法处理正偏态资料的合理性,并对多份实际资料作了拟合尝试.拟合结果表明:实际频数分布与理论频数分布的拟合优度好,直线化处理后的线性程度高,可提供的信息量丰富,对不同偏态程度的资料适应性强。在医学研究中,正偏态资料颇为多见,因而该法在处理这类资料时具有广泛应用价值。
Usually, the data of positive skewness distribution is processed by geometric mean method, but for those data that can not become normal distribution after logarithmic transformation, there is no theoretical basis for doing so; however, the use of percentile method can not be sufficient. Using the information contained within the data, this paper theoretically clarifies the rationality of using Weibull distribution method to deal with positive skewness data, and makes fitting experiments on multiple actual data. The fitting results show that the actual frequency distribution and theoretical frequency The goodness of fit of the distribution is good, and the degree of linearity after linearization is high, the amount of information available is abundant, and the data of different degrees of skewness are highly adaptable. In medical research, positive skewness data is quite common, so the law has wide application value when dealing with such data.