4
$\begingroup$

I am having trouble reading my qqplots. I want to asses normality graphically before I carry out a test like shaparo test(my sample is small (<50)) . So I have two question is the qqplot better than a box plot for assessing normality and if so why.

In my analyses I carried plotted the qqplot and the points lye more or less along the line therefore we could not reject the normality assumption, but I wasn't quite sure as I found the plot hard to read so from the box plot it is very slightly skewed to the left. Any feed back on this would be much appreciated.Box Plot v QQ plot

$\endgroup$

2 Answers 2

2
$\begingroup$

With a QQ plot you show all the data points; with a box plot you lose a substantial amount of detail. So yes, QQ-plot is better for looking at normality. Just looking at the graph I would say that your variable is a bit heavy tailed, but for many applications it will be close enough to being normal.

$\endgroup$
2
$\begingroup$

The problem with the box plot is you have no reference distribution against which to compare your sample, so little way of assessing normality. The Q-Q plot is much better suited to this purpose. Or you could compare a histogram of your sample with a normal density with the same mean and variance (here is a thread from Stack Overflow about this). This would be arguably easier to understand than the Q-Q plot.

$\endgroup$
1
  • $\begingroup$ Very good point with the histogram, however we have covered box plots and we are on qq-plots and and I was just wondering what the differences were in them when assessing normality and I was hoping to get some feed back on my qq-plot. $\endgroup$ Mar 2, 2016 at 14:59

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.