causamed

Implementation of Tyler Vanderweele's methods for mediation

####Everything pulled straight from the book "Explanation in Causal Inference; Methods for mediation and interaction" by Tyler Vanderweele (2015)

Overview and changelog:

Currently implements

1. Regression methods for continuous mediator/outcome (section 3.1.1)
   
2. IPTW-based method for scenario with exposure-induced M->Y confounding (section 5.3.1)
   
3. G-estimation based method for joint effect of multiple mediators (section 5.3.3)
   
4. Calculation of random interventional analogues to direct and indirect effects (section 5.4.2)
   
5. version 0.1.0 : Structural means models to deal with exposure-induced confounding of M->Y given a continuous exposure or mediator (section 5.3.5)
   
6. version 0.1.1 : Handling of mice objects to add to resampler (slow but more stable than pooling from few imputed datasets)
   
   

To do (not much work)
5. Structural means models to deal with exposure-induced confounding of M->Y given a continuous exposure or mediator (section 5.3.5)
6. Handling of mice objects to add to resampler (slow but more stable than pooling from few imputed datasets)
7. Extension of methods to categorical and continuous outcomes where not already the case
8. Regression-based methods for multiple confounders
9. Check random interventional analogue for categorical exposure case
10. Add methods to summary() generic function

To do (lots of work)
10. Implementation of sensitivity analyses for regression-based methods
11. More sensitivity analyses in settings where NDE/NIE not identified
12. Handle non-parametric/semi-parametric approachs to propensity score estimation for weighting-based approaches