Statalist The Stata Listserver

[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: Population-pyramid-like graph - positive barlabels on the left?

From   "Nick Cox" <>
To   <>
Subject   st: RE: Population-pyramid-like graph - positive barlabels on the left?
Date   Fri, 21 Jul 2006 12:50:12 +0100

It's not the answer you want, but I really don't understand
why these pyramids are so popular, or even why their key 
limitation is not mentioned more frequently. It must be
some sort of social feedback loop, as with pie charts, that
within certain groups pyramids are familiar, therefore expected, 
and vice versa. 

The aim of your graph is presumably to compare male and female, 
which comes down to either a difference or a ratio. A pyramid 
obliges the reader mentally to pick up one bar and lay it down 
on top of its neighbour. I don't believe that many of us are 
very efficient at doing that in our heads. Also, I guess that 
whether males are just a bit higher than females, or vice versa, 
and that even small differences or ratios very near 1 could be of great
interest. (Contrasts for breast or prostate cancers presumably wouldn't
surprise anyone.) But how easy is it to _decode_ such contrasts 
from pyramid representations? 

Otherwise put, your desire to show bar numbers as well is 
a tacit admission that the graph form is inefficient at 
showing patterns that can be decoded. 

You would, I assert rashly, be far better off with -graph dot-,  
which encourages genuine quantitative comparison.  


Gawrich Stefan
> with cancer registry data I want to draw the sex-specific 
> distribution of
> cancer localisations on a population-pyramid-like diagram.
> The difference to the population pyramid is that I have localisations
> instead of age groups on the vertical axis, which differ by 
> sex. So I have
> different localisations to the left (women) and to the right 
> (men). The
> localisations are sorted by frequency and to obtain a 
> homogeneous graph I
> use the same number of bars/localisations on both sides.
> I tried different graphing options and ended up using two "gr 
> hbar" graphs
> combined so far. 
> Whats missing is that I need to label the bars (abs. or rel. 
> frequency) and
> get negative values on the left side. Is there a way to 
> obtain all positive
> bar labels (just like the options for all positive axis labels)? 

*   For searches and help try:

© Copyright 1996–2021 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index