Tue Sep 4 23:05:00 CEST 2012
How to make a good density plot?
I know of only two ways:
- Histogram: tuning is the interval resolution. Trade off: more
space resolution gives less density resolution.
- Kernel method: Each discrete event is spread out in space. Basic
"tuning" parameter is the size of the kernel. Trade off: more space
resolution 9smaller kernel size) gives a more bumpy density curve.
I can think of some mods:
- Adaptive kernel? (E.g. for financial data?) Decrease kernel size
in high-density regions.
It would be good to figure out a good way to do this in Haskell +
Gnuplot. Probably needs the preprocessing (events -> curve) in
outside of Gnuplot.