1: These examples are drawn from both goodness-of-fit and regression traditions. Goodness-of-fit
tests look for specific distributions in data, such as testing a variable for normality.
Regression fit looks specifically of an association that is a straight line. The underlying
principle is essentially the same in both cases: we are looking for patterns in data.
[return]
2: You can ask for both skewness and kurtosis statistics in SAS in the PROC UNIVARIATE
module among others, see the code “Code07b Univariate descriptives” for an example.
[return]
3: Cut-offs for skewness are debatable – some authors suggest +-1, others +-2.
[return]
4: However, the more sophisticated tests have their own dangers, as discussed in the
next chapter. In the case of normality testing, in fact, these tests are usually not
used as they are highly flawed because the size of the sample can make them ”see”
non-normality even for data that is close to a perfect normal line.
[return]