Stochastic Treatment TMLE

[pzivich]

This is another valuable addition to TMLE (that I also need as part of a project I am working on). Essentially, this would allow more complex treatments than treat-all vs. treat-none, similar to custom treatments in the g-formula.

What it does is shift the probability of A distribution. However, the single-step convergence of TMLE is no longer valid. I would need to iteratively estimate Q* until it epsilon converges to 0. This should be easy enough. After convergence, follows the remainder of the TMLE procedure

Likely best if I make this separate from TMLE

Starting points;
https://github.com/tlverse/tmle3shift
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4117410/#SD1

[pzivich]

A heuristic for how to approach IPW estimation for stochastic treatments. (1) Weight the population to the all vs none comparisons (standard procedure), then (2) reweight again to the intervention targeted.

[pzivich]

General reminder to myself: stochastic treatment TMLE is no longer a one-step fit. There has to be an interative process of estimating Q* until convergence criteria is met. This means repeating the targeting step several times. while loop for this with either a convergence criteria or max_iters criteria should be the solution

[pzivich]

For comparisons, I can use R'stmlenet which defaults to the standard TMLE when no network is provided

[pzivich]

Added as part of 0.8.2