TODO.org

Do this now

Checks (Thomas)

Check if the functionality works in all these cases with nonstandard names for censoring and competing risks

• with competing risks and censoring
• without competing risks
• without censoring
• without competing risks and without censoring

Output (Puriya)

Improve summary and plot functions

Flaskehalse i stor data (Hannah)

• remove long and list of wide data

Positivity (Kathrine)

Long to Wide for start stop variables (Christian)

• discretization of register data
• examples of non-chronical variables
• effect of different interval lengths on wide format (used drug x% of the time)

Do this soon

Check if outcome is constant before running anything

allow subsets to define Strata as a vector of values (requires that the variable is in the prepare_rtmle_data)

Related literature (check these out)

Trimming (instead of g-bounds) time dependent? diaz + post-doc

SATE survivor average treatment effect

Describe extensions

Find alternative estimators and corresponding R-packages

Pooled logistic regression for g-part (censoring/propensity)

How to set up the data: multiple lines per subject?

How to summarize the history: for sum(bar L) the range of the variable increases with time

What is the correct loss function? patients should be similar wrt outcome risk not propensity!?

Import various Hely-programs

https://github.com/helenecharlotte

• Dropin
• TMLE for baseline treatment and baseline covariate event history analysis
• TMLE for recurrent events (repeated measures)
• continuous time longitudinal causal inference