I am looking for (any) pointers on how to approach the following abstract problem. Not: my statistics background is very limited, so I might very well be missing something obvious.

We have subjects each being subjected to a sequence of procedures. Different subjects might undergo different treatments, in a different order, but all the treatments come from a common catalogue (which can contain hundreds of different treatments). Furthermore, a treatment can be parametrised by one or more continuous parameters (e.g. intensity and duration). We have measurements for subjects state before and after the treatments — these are vectors of continuous values. We can furthermore assume that variables are normally distributed.

The question is: how can one find the sequences (and parameterisation) of treatments that show the strongest effect in maximising/minimizing a selected subset of state. E.g.: "in order to improve score X, one should do treatment A(a=0.5, b=0.7) followed by C(a=1.3) followed by A(a=0.2, b=1)"

My initial idea was to reformulate all this as a linear regression model, but we end up with hundreds of variables, some of them having interactions, so I am not sure that its the way to go...


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.