Given Allan's charming start, it seems churlish
to dissent.
But densities of continuous variables can have any
non-negative value consistent with their integrating to 1
over their support. This is not a point for discussion: it
is standard probability theory arising directly from
integral calculus. A very clear example is embedded
in the manuals, at [R] kdensity I think. It follows also
from the definition that changing the binning can change
the empirical densities shown. It would be alarming if this were
not so.
Nick
[email protected]
Allan Reese
> Nick Cox is right:
> Allan's complaining about perceived perversity, but I am not
> clear what he
> would regard as good behaviour.
>
> Jann Ben is also right, and missed my point:
>
> The purpose of a histogram is to make visible the shape of a density.
> It is therefore natural to report the y-axis in terms of a density.
> -------------------------------------------------------------------
>
> I can see this is a case of computers do what you tell them,
> but I'm not happy with it.
> As was also pointed out, there are alternatives and I'm the
> first to agree that users should
> look at and understand the graphs they draw.
>
> But there seems something inherently wrong in calling a scale
> "density" when it can have
> values greater than 1. Nor is it obvious to me what point
> there is having a scale when the
> values may change with the number of bins. I must remember
> to add "frac" every time, like
> I add ylab(,angle(0)) as a reflex ;-)
*
* 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/
