information measurement of two data sets - how to start? (Apr/06/2008 )
I have two heatmaps of two different physiological events but each event in dependence of the same transformation; one heatmap is color-coded by white/blue, the other by white/red; strong color codes for strong event. If I merge both heatmaps, I get a color-code of white/blue/violet/red.
In the case of the overlay, a low violet, for instance, stands for two different low events,
white for no event at all,
strong blue, only one event strongly happens, the other not,
and so on.
I try to find out how close the two different data sets depend on each other. There is very likely no linearity of synergism but how to start to find the possible interaction rules? Any mathematics or concepts to apply?
why don't you apply a range of numbers that determines the heat intensity? i.e. 0 for the lowest value, or the mean value, with the possibility of having negative numbers for blue (the strongest blue could be -10, depending on the nature of your experiment), and to the strongest red +10, that would create an interval with all the variants of colors, then you should be able to do statistics on that.
You already have a scaled set of vectors. Hue, saturation and intensity. You could use these values in a nice bit of vector math.
For an idea have a look at: residual colours: a proposal for aminochromography, W. Taylor, 1997, Prot Eng
Matlab has some libraries for this sort of thing, as does R.