I've noticed an extremely frustrating, not to mention misleading, aspect of
Excel 2K. If you have a column chart with a series whose values are all
considerably greater than zero, the "autoscale" feature will by default cut
off the floor of the graph at some value greater than zero. Charting
textbooks abound with comments as to how this is a very misleading style,
because it tends to exaggerate differences between points which might
actually be relatively quite small. Someone who's not looking at the chart
with a sharp eye might conclude that 2 points differed significantly when
in fact they were pretty close.

I was able to change the chart to zero-origin, but the process was fairly
long and in any case it's irritating to have to do this on every chart. Is
there a way to set options or registry settings such that it will always
default to zero-axis for any column (or similar type) chart regardless of
data values (the one exception, of course, being if you were plotting on a
log scale)? I'd like to do that and solve the problem once and for all.

In a larger context, how could Microsoft think their approach was an
intelligent default? Given that the literature emphasizes the enormous
risks with non-zero-origin charting, and how easy it is for such charts to
be badly misleading, what thought process led Microsoft to do it anyway?
It's especially bad in view of the apparent fact that trying to change this
is pretty obscure and many people might not figure it out (or indeed even
notice in the first place). So people might being end up misled by their
own charts, not to mention inadvertently misleading others - or certainly
ending up in frustration when the way their charts turn out isn't how
they'd like them to display.

Meanwhile, for myself it's also clear to me that in addition to Excel I
need a heavier-duty charting package as well. I routinely need to make true
3-D surface plots, and I'd also like to be able to plot a range of standard
functions, do complex curve fits, do charts with real and imaginary axes,
using various coordinate systems (e.g. cartesian, polar, hyperbolic, etc).
and in general get the power of full mathematical analysis. However - and
this is the key point - I need to be able to do this *without* having to
write out or calculate equations manually. Programs like Matlab are very
competent with mathematical calculations and graphing, but for simple
situations the setup time makes it not worth the effort. There are times
when I want to do that, and for these applications Matlab is a good tool,
but for times when I need to pound out a quick chart it's just a lot of
work. Does anybody have recommendations for good charting packages for the
type of activities I'm describing?

--
Alex Rast

(remove d., .7, not, and .NOSPAM to reply)