Iterating Data Analysis

[dabayliss]

First a caveat - my knowledge of medicine is minimal - so it is plausible I have misunderstood some of your words. Notwithstanding:

1) The only question here is whether or not you know your grouping criteria apriori. If you do then a simple project where you use a MAP to convert your data into categoric variables will work fine.
If you do not know the grouping apriori then the Discretize routines inside ML allows you to categorize the various variables into either percentiles or evenly across a range.

2) This is the bit i am least likely to understand. In order to compute survival rate (as I understand it) - you need to know who is entering (through diagnosis) and who is leaving (through death) the population. That is what my first and last is computing. If you have those two I -think- you can compute the survival rate.

3) This is a straightforward clustering exercise. The agglomerative clustering piece of the ML libraries perform both the distance metric (we have about half a dozen) and the hierarchical clustering for you.

4) The above all work at extreme scale ...

[pyrannis]

Thanks dabayliss,

I will give that a shot and let you know how it works out.

Stetson