[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: RE: quirk: xlab overrides xscale in Stata8
> Just upgraded to Stata 8, latest updates, Windows XP. A
> quirk discovered
> in graphics. Stata8 does not respect the "xscale" option
> when it figures
> out the "xlabels", and then "xlabels" takes precedence over
> "xscale" in
> the drawing of the axis. It appears that "xscale" can only
> be used to
> expand the xaxis to the right, not truncate it (even though you are
> within the limits of the data). You have to change both
> xscale and xlab
> to achieve the desired effect. An example:
> sysuse auto
> *range of weight is 1760 ro 4840
> summ weight
> *default plot, x scale ranges from 1760 to 5000:
> scatter turn weight
> *plot with x scale expanded to 6000, no problem:
> scatter turn weight, xsc(r(0,6000))
> *plot with x scale restricted to the maximum value of the
> data, does NOT
> scatter turn weight, xsc(r(0,4840))
> *you have to also change the xlab option, then xscale takes effect.
> scatter turn weight, xsc(r(0,4840)) xlab(2000 4840)
> This is somewhat counter-intuitive.
This is not really a quirk, but it is a change.
In general, one change from 7 to 8 is that "nice round" labels
(as formerly obtained by -xlabel- and/or -ylabel- without
arguments) are now the default. Previously, the default
was of course labelling based on data extremes. We could
argue about whether that is a better default, but my
own view is that it is.
In this example, the default labelling of -turn- is
2000 3000 4000 5000 and as Jan notes only by
spelling out that you want something else
can you change the labels.
As before, the scale limits set -- now through
r(min, max) -- act only to extend the axes from
what the graph would be otherwise. In this
case, only by changing the labelling explicitly
does the "otherwise" change.
* For searches and help try: