I have a set of data (positive real numbers < 1) in five categories. My aim is to show that the data in the last category is bigger than the other categories for a range of examples (the data set), but the last category isn't necessarily more important than the others.

I thought the best thing is to average each category and then plot the graph of it which works. But I feel this is too simple and is it good if I look at the standard deviation of each category to show that there isn't too much difference and the average-taking is valid?

Also, I was going to use a scatter plot, is this OK? What would you recommend?

  • $\begingroup$ Some sort of plot would be sensible to explore the data. It is difficult to comment on your other statements without more information such as what "more important" might mean. $\endgroup$
    – Henry
    May 8, 2011 at 15:36

1 Answer 1


For plots, it depends on the size of your data set. With relatively small N, I would do a parallel strip plot, which is more-or-less a scatter plot. I might add jitter to the category variable or change the plotting symbol, or make it transparent, depending on N.

For a test, if you want to test which category has the biggest mean, then ANOVA with an appropriate contrast would be the way to go. For biggest median, you could use quantile regression.


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.