0

background

We are running an operation using the variables in the following table to result in an outcome under the ATC success rate % column:

enter image description here

(note: this table is only a sample.. the number of permutations can be very large, ie (number of providers) * (number of machine categories) * (proxy providers) * (proxy subnets) * (websites), see here).

problem

The point of tabulating this data is to find a pattern amongst these variables, for example: the combination of P1 and MC1 and PP1 and PS1 and w1 yields the best result.. but the causal relationship isn't clear, ie what is the variable that's causing the best result? is it the proxy provider or the website?

enter image description here

Question

How can this causal relationship be derived, mathematically speaking?

Community
  • 1
abbood
  • 101
  • 2
    Anything trying to demonstrate "causality" has to show it is not just "correlation", usually by undertaking controlled experiments rather than observation of what is happening anyway. – Henry Apr 12 '17 at 11:45
  • so in the context of the above, what would constitute a "controlled" experiment? ie keeping everything constant and changing a single variable at a time? (ie to avoid confounding variables etc) obviously the data has to be statistically significant for the outcome to be credible right? – abbood Apr 12 '17 at 11:47

0 Answers0