@JoshMonkey wrote:
I know this isn't a lv problem but more of a conceptual and application of statistics and mathematics but can anyone suggest a method to enable me to classify a data set into "best fit groups" (number of groups unknown and to be determined)? My understanding is that to do so would mean that I need to test all the possible combination of data ranges and find the one with the least variance, is that correct?
I suppose that you are referring to your sphere problem (in the vision board).
When dealing with multivariate data, there are some automatic classification methods, that need more involved computation than a mere variance comparison. May be your first step could be to do a Principal Component Analysis (PCA) and see what you get.
KevinC did a really good job (that deserves some more stars...) and from what I have seen on the pict that was provided, mainly four groups (red, green, orange and yellow) can be easily distinguished with the hue intensity and index. The info is given by indexes 3, 5, 7. When the max value is higher than 0.8, you have a "pure" color (index = 3 -> red; index = -> orange; index = 7 -> green), when the max value is under 0.8, you have a mixed color (here : only yellow).
Please share your findings, and ask if you need help with PCA.
Hope that helps !
CC
Chilly Charly (aka CC)
