How to limit y-axis on box plot to exclude extreme outliers

[Stephanie Cleland]

Hi,

I am trying to create a series of box plots for a set of data, but am running into some issues. What I want to do is include extreme outliers in the calculation for my boxplot (e.g. so they are included in the calculation of Q1, median, Q3, etc.) but I don't want outliers above a certain range displayed on my graph. The reasoning behind this is some outliers are so extreme that it makes it impossible to see the actual box plot.

The current code I am using to generate my boxplot is:

graph box Outdoor Indoor, ylabel(0[20000]100000)

And the outputted graph I am getting from this code is:


So what I would like to do is limit the maximum value displayed on the y-axis to 100000 (as I attempted to do with ylabel).

Thanks so much in advance,

Stephanie

[Carole J. Wilson]

You could specify the nooutsides option which would not plot any of the outside values. Alternatively, you could try a log scale yscale(log) if that makes sense.

[Nick Cox]

yscale(log) may not do what you think it does. I've seen people bitten or at least confused by this, so the point is expanded in excruciating detail at http://www.stata.com/support/faqs/gr...ithmic-scales/

My own take on data like this is that you need to think on a different scale, perhaps logarithmic or cube root or square root. Once you do that, the box plot will be much more informative and won't need radical surgery.

[Carole J. Wilson]

Nick Cox Thanks for the reminder about box plots and non-linear transformations! That's a nice FAQ.