Reporting the skewness statistic in text

#1
I am writing up my dissertation and need to report a skewness statistic of one of my variables because it is just outside the accepted range to be considered normally distributed, but I have no idea how to quote in in the text, APA (American Psychological Society) style.
Does anyone know??!! Thanks.:wave:
 

Dragan

Super Moderator
#2
I am writing up my dissertation and need to report a skewness statistic of one of my variables because it is just outside the accepted range to be considered normally distributed, but I have no idea how to quote in in the text, APA (American Psychological Society) style.
Does anyone know??!! Thanks.:wave:
I'm wondering why you don't conduct a usual test of normality using one of the common statistics available?...e.g. Kolmogorov-Smirnov or Shipiro-Wilks test and report the statistic and p-value. These test are available in SPSS and other software packages.

Further, I don't understand how you can only consider the skewness of a variable in the context of testing for normality without at least considering the kurtosis as well.

Consider the following:

1. Normal distribution has skewness = 0 and kurtosis = 0.
2. Uniform distribution has skewness= 0 and kurtosis = -1.2
3. Bernoulli distribution has skewness= 0 and kurtosis = -2
4. Logistic distribution has skewness = 0 and kutosis = +1.2
5. Double Exponentional dist. has skew =0 and kurtosis = +3.0

2,...,5 are all symmetric non-normal distributions i.e. skew = 0, with kurtosis equal to something other than zero.

In other words, varying the degree of kurtosis has an impact on the shape of a distribution especially in terms of the tail-weight.
 
#3
Hey guys,
I'm in a similar position to the first guy. Dragan's right about considering kurtosis as well as skewness - but does anyone know the actual APA guidelines on reporting violations of normality assumptions? I haven't had any luck finding them, and stats textbooks always seem to give examples for normally-distributed data.
 
#4
I would suggest referring to Tabachnick & Fidell 'Using Multivariate Statistics' textbook, that helped me out alot and gives examples on how to report 'non-normal' skewness and kurtosis and transformations etc.
 

CB

Super Moderator
#5
I can't find a guideline for skewness/kurtosis in specific, but I'd assume that one can report these similarily to how you'd report a mean or standard deviation.. i.e. (skewness = 0.71, kurtosis = 0.75). I wouldn't italicise "skewness" or "kurtosis" though, since they aren't abbreviations. 2 dp's probably best.

http://my.ilstu.edu/~jhkahn/apastats.html (reference for how to format mean/sd).

APA stands for American Psychological Association, btw :)
 
#6
you report it like this:

D(91) = 0.12, p < 0.05

(91) is df (on SPSS printout)

= 0.12 is your statistic (on SPSS printout)

p< 0.05 is your alpha