Im new to data science, and was wondering if someone could point me in the right direction to start understanding how to perform the correct statistical analysis/decomposition/modeling analysis of the following data set.
I have a number of continuous variables that I can add or not add to a control sample (which has no additives) to change the output variables by some amount relative to the control's output values (example dataset picture attached).
How can I correlate each input to the output and use this to build a model that will help me predict the expected output given the addition of some new ratio of additives A, B, C, D, E, F, G at amounts a*, b*, c*, d*, e*, f*, g*. A-G are not necessarily independent.
In the real data set, I have ~400 sample data points for ~30 input variables, and ~20 output variables, so machine learning will not help as this is most definitely intractable. Further, this is a biological system so building the underlying network itself would be worthy of a PhD, an amount of work that Im trying to figure out how to avoid.
Interesting problem to be working on Nikhil Goel . If they are not necessarily independent I'd think you'd want to consider a mixed model (accounting for the co-variance).
Interesting, thank you Edward G. Brown . Very helpful.
How would your response change if I said Im also interested in the data set where A-G dont necessarily have independent effects on the output variables, not that the variables themselves are potentially dependent on each other?
Ahh, in that case it sounds like you’d have mediating and/or moderating effects. Could be bery interesting. I think David Kenny has work on that topic which could be helpful.
Perfect, this is exactly what I was looking for! Moderating and dyadic effects are going to be highly relevant, thank you for the great insight! This is going to be very interesting indeed...
- Badges
- Science topic