Create histogram of continuous data with two bins to catch outliers on the tails

[josh scott]

I would like to create a histogram that charts the continuous values of a particular variable. Most of the observations range from 2 to 5. I would like to have the main bins with a width of 1 and range from 0 to 7. However, I would also like two bookend bins that cover 0 to min(variable) and 7 to max(variable). When I try to do this, I always get errors with the way I use bins, or if I try to use the discrete() or cut() options with addplot().

Here is a data example of the variable:

* Example generated by -dataex-. For more info, type help dataex
clear
input float logA
   3.595926
  1.4829882
   2.418526
   2.232657
  2.2168064
   9.851233
  1.9928473
  2.0715392
   2.878051
   3.741612
  2.5099146
   3.546768
   2.815486
   1.762309
    2.36012
 30.12
    1.93512
  3.0619204

end

How can I achieve this in an effective manner?

Thank you in advance!

[Maarten Buis]

Maybe you are looking for eqprhistogram by Nick Cox? To get it type in Stata ssc install eqprhistogram. Applying it to your example, I got:

eqprhistogram logA, bin(5)

[Nick Cox]

This is an intriguing example, although with an anonymous name, no context, and quite possibly only a tiny subsample it would be easy to guess wrong.

Something called logA is quite possibly the logarithm of A and if there are high outliers on a log scale they must be massively hign outliers on the original scale.

I note further only that if you use not logarithms but reciprocals the transformed data are close to a uniform distribution.

[josh scott]

Thank you Maarten Buis! I will take a look at this.

Nick, this is a great point. I actually do have the reciprocal as well, but I wanted to look at both. 95% of the observations would lie within the seven bins with a width of 1 and only 5% above that. eqprhistogram is a great option, but am also curious about how to make this upper bin. For my preliminary idea of this histogram, I just wanted to show how many observations fall in each bin, and show the 5% above 7, but then further dig in to the range of the outliers elsewhere if necessary.