https://bbs.pinggu.org/thread-1100084-1-1.html

谢谢版主，sktest的说明我在提问之前也仔细看过，原本以为只是做正态性的检验。后来发现问题实际上是在对Pr(Skewness)和Pr(Kurtosis)的理解上，关键是不明白他们代表的就是p值。我搜索Pr(Kurtosis)，找到了一篇材料，其中对sktest有更为细致的说明：For sktest, the null hypothesis is that the skewness is 0, kurtosis is 3 (the same as the normal).
The values labeled Pr(Skewness), Pr(Kurtosis), and Prob>chi2 are p-values.
A low p-value corresponds to rejecting the null hypothesis.
In this case, mpg has a rather high skewness, near 1, and we reject the null hypothesis that the values of mpg come from a normal distribution.
2# sungmoo

这里就看“峰度”采用哪种定义。

*stata计算偏度与峰度的方式

egen sk=skew(x)
*等价于
egen m=mean(x)
egen s3=mean((x-m)^3)
egen s2=mean((x-m)^2)
g sk=s3/s2^1.5

egen ku=kurt(x)
*等价于
egen m=mean(x)
egen s4=mean((x-m)^4)
egen s2=mean((x-m)^2)
g ku=s4/s2^2