[moneydance] Graphing Awesomeness

[Ric Werme]

Edward Reid wrote:
> At 21:32 01/03/07 -0700, Hans Derycke wrote:
> >- the y-axis of graphs is now shifted to a value closer to the lowest
> >y-value, instead of being stuck at zero.
> 
> Those with some experience in statistics and presentation consider this a 
> Very Bad Idea, because it emphasizes changes which are of little 
> significance. Without even having looked at it, this comment leads me to 
> assign ten demerits to the new graphs.

I completely understand your stance, but prefer the "cropped" display.
I also prefer a logarithmic scale on the Y axis for some graphs, and
those don't have a $0 level.  However, that's another story.

> >For instance, if your net
> >worth graph varies between 1,000,000 and 1,020,000, you'll see every
> >painful nuance of that variation, instead of a largely blank screen
> >with a wiggle at the top
> 
> But in this example, the variation is +/- 1%. In fact that IS just a wiggle 
> at the top of a largely unchanged situation. Making that +/-1% variation 
> look like an earthquake is bad presentation.

http://www.edwardtufte.com argues that graphs should have a maximum slope
of about +/- 45 degrees.  Just a guideline, and I'm sure one he'd derate for
financial data, but flat lines and earthquakes both hide information
from the skilled reader and hence may be poor presentations.

Given that mutual fund prospectuses are quick to point out that past
performance history is not a guaranteed predictor of future performance,
perhaps people should use graphs at all!  Of course, the counter cliche is
that those who forget the lessons of history are doomed to repeat the
mistakes of the past, then perhaps the only history range to display
starts with the first non-zero value.  Hmm, my Vanguard Windsor Fund IRA
looks pretty good over the last couple of years, even with a display that
starts in 1980 and displays a $0 line

I do agree that your +/- 1% example is a poor graph to dwell on, but at
least we now have _some_ control by going further back to pick up a
wider range.  One thing you didn't mention is that it's virtually
impossible to compare graphs where the Y range is automagically
chosen.  Ultimately control over everything is good, except for ease
of use.

One thing that might work without user input and be easy to implement is
to ensure the Y range is at 10% of the max (or min or mean).  It's a bit
of a special case, but my barometer graph at
http://home.comcast.net/~ewerme/wx/current.htm always displays between
29.40" and 30.60" and extends either limit to handle extreme weather.  I
despise barometer graphs that expand the trace to fit the vertical space
available ala your +/- 1% graph.  When the pressure is not
changing much, you have to read the Y scale to see if the pressure is
high or low.  On my graph it's obvious.  None of that really applies to
Moneydance, (except that you might argue I should show 0.00"-30.60"),
but my windspeed and rainfall graphs show at least 0-10 mph and 0-0.1"
which should meet with approval from most statisticians.

At any rate, I greatly prefer the non-zero base graphs.  I wouldn't say
they hold awesomeness, but will say "Yay."  :-)

	-Ric Werme