Hi there
It seems to me there's a fast growing base of support for histograms with
varying widths in Stata. This is just to add my vote in favor of
implementing such an option in Stata 9, and in further ado files in the
meantime.
Nick Cox's arguments against are: 1. with the way data are these days, few
people will need histograms with varying widths, and 2. it's a slippery
slope to infinitely large or small supports.
My responses are:
1. There are data sets that allow only varying widths in the support, and
not all of them are obscure. Take for example the public use Statistics of
Income (SOI) data published by the United States' Internal Revenue Service,
which gives frequencies in unequal intervals of taxable income in the U.S.
Moreover, there are many instances where unequal widths in the support make
natural sense. Consider if one were to examine frequency of drug use by
age, one may want to examine frequencies by broad age categories that are
clinically meaningful but not of equal interval. The age ranges may be:
young (0-12), adolescent (12-17), young adult (18-24), ..., mature adult but
not retired (50-65), etc...
2. Let the user decide where to draw the faux-closed interval for what is
logically an infinite continuum. Us users are used to making such
decisions, and we usually know where we can ask if we need help deciding.
Often the subject discipline will give guidelines based on precedent. This
is, after all, how top-codes are often used in reporting income in many
survey data sets.
In fact, my case is a general plea to allow for greater user power in the
histogram command.
Thanks
Alex Cowell
Alexander J. Cowell, Ph.D.
Economist
Behavioral Health Economics Program
Research Triangle Institute
3040 Cornwallis Rd
PO Box 12194
Research Triangle Park
NC 27709-2194
email: [email protected]
phone: 919 541 8754
fax: 919 541 6683
Nick Cox wrote:
1. Empirical. You will see histograms with unequal widths particularly in
older books and papers, and the reason was that data for them came already
grouped in such classes. There's an example in Snedecor and Cochran's
venerable text. That seems far less common today when more and more data
sets are available in raw, ungrouped form, modulo confidentiality
constraints. I don't see people asking for this often on Statalist, and one
good reason for this being low down in priority is that it is practice
rarely needed.
2. The "slippery slope question": if unequal widths are supported, then next
in line is the question of support for a histogram with a class which
extends from large positive number to infinity and/or a class which extends
from a large negative number to minus infinity. Even quite what you _should_
draw then seems to me an open question (pun intended).
Richard Goldstein wrote:
I very strongly disagree with Nick's conclusion here (even taking it as
somewhat tongue-in-cheek): For any graphic command that has an option such
as bin (histogram), bwidth (lowess), width (kdensity), I would very much
like to see dynamic graphics -- i.e., a slider such that I can change, e.g.,
the number of bins in real time and see what kind of difference it makes to
the graph. Would anyone else like to see something like this? Rich Goldstein
Marcello Pagano wrote:
As I have said before, I would very much like, with Allan Reese, an option
for equi-probability histograms. I think that this is especially useful when
thinking of the histogram as an estimator of an underlying density function
of a continuous variable. I do not see a strong argument, other than
tradition (laziness?), for being constrained to histograms with equi-spaced
bins.
m.p.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
