0
$\begingroup$

I currently have a vector containing values which I wish to consider the distribution of. I am aware that when using a qqnorm() plot, it is possible to see that data is normally distributed if it closely approximates a straight line.

However, the qqnorm() plot for my data looks something like this:

qqnorm

My understanding of r, and of this function, is very limited. However, from searching online I suspect that this graph may tell me that my data is normally distributed with skew. How might I check whether this is indeed the case?

$\endgroup$
1
  • 3
    $\begingroup$ "Normally distributed with skew" is a contradiction in terms: no Normal distribution is skewed. Please explain what you're trying to ask. $\endgroup$
    – whuber
    Mar 17, 2018 at 18:26

1 Answer 1

3
$\begingroup$
  • The Q-Q plot indeed shows that your data are skewed; the values in the lower tail are larger (closer to the median) than expected from a Normal distribution (light left tail), while the values in the upper tail are also larger (but thus farther from the median) than expected (heavy right tail).
  • There's not really any such thing (that I've ever heard) as "normally distributed with skew"; I would say these data are "skewed" or "right-skewed" (e.g. see here).
  • If you want to get more of a feel for your data, it may help to draw a histogram; it's harder to detect subtle differences from Normality, but it's easier to understand what's going on, at least until you get the hang of reading Q-Q plots.

    enter image description here

  • you're asking whether you can "check whether this is indeed the case". It sounds like you're looking for a statistical test of skewness. While there are tests (such as the Jarque-Bera test, available for example in the tseries R package) that use skewness and kurtosis to test for deviation from Normality, I'm not aware off the top of my head about tests of skewness alone. That said, I would ask why you want to do such a statistical test: what question are you ultimately trying to answer? Many statisticians feel that statistical tests of Normality are mostly useless; similar arguments would apply to a test to reject the null hypothesis of symmetry (non-skewedness) ...

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.