Plotting Medians Vs Means: non-parametric data

[rosiedrop]

Hi,

I'm a PhD student who has been spending a lot of time lately analysing data (as seems to norm!) - a lot of my data sets fail normality tests, although show significant differences in the rank of means.

My question, however, is what is the best way to represent such data? I originally prepared graphs of the mean, but then worried that because the data is skewed, that was wrong, so I have also made a graph of the median. I considered boxplot, but they don't seem to be used at all in the relevant literature...

Any thoughts on this would be gratefully received!

Thanks-you.


(I also have a lot of zeros in my data, but all the talk of poisson distribution etc confuses me - I use SigmaStat and can't find which test will help me!)

 

[bugman]

I'd recommend notched boxplots for comparisons between groups. Or try plotting median values plus/minus a given percentile.
What area are you working in and what do others in your field use? If they don't use boxplots?

 

[gianmarco]

Hi!

Info about box-plot here

Regards,
gm