TODO list

[jialul]

Causal ML updates

General issues

• check package environment with Windows
• write up the documentations
• add examples (micro-finance/star)
• flow chart for the package #28
• Estimate ITR with observational data and evaluate with RCT #38

Specific issues

Documentation

• finish documentations for test_itr
• Roxygenize all the function and data
• Simplify the examples

Functions

• check whether return outputs are consistent across all functions
• check the SVM function
• write the functions for binary outcomes #8
• add GATE to the qoi #9
• add functions for the statistical tests #17
• add infrastructure for training under sample splitting #20
• separate training/testing with compute qoi #26
• Incorporate different types of meta-learners #24
• add het.test and hetcv.test to the package #36

Testing

• use Michael's clinical trials data to test functions for binary outcomes #15
• add testhat tests #14
• Run CRAN checks with rhub

Output

• create a summary(fit) page where we display some summary statistics (perhaps some PAPEs, fitting metrics and etc)
• add summary of the plot_aupec output
• allow summary() function to show all the output (PAPE, PAPEp, PAPD, AUPEC) #10
• allow the function to export p-values #11
• skip PAPDp in the summary output if the user only chooses one algorithm #18
• allow users to estimate the ITRs with their own models #27
• Check the sample splitting plot #30
• If only one alg is selected, do not print PAPDp summary statistics.

Plotting

• check the variance for AUPEC

Other tasks

• incorporate Python functions and call them into R #12
• create website for the package #13
• Check the tuning parameters for caret and SuperLearner

[kosukeimai]

What about adding some tests via testthat?

[jialul]

@kosukeimai will do!

[jialul]

May need to increase computation efficiency: http://adv-r.had.co.nz/Profiling.html

[xiaolong-y]

Could be useful applications for our paper or demo on the package website: https://github.com/gsbDBI/ExperimentData

[xiaolong-y]

May need to increase computation efficiency: http://adv-r.had.co.nz/Profiling.html

Definitely good idea! Perhaps we can look into this once we finished writing all the functions planned for the current stage?

[xiaolong-y]

@jialul summary now shows statistics for GATEcv as well! Some new columns are upper and lower bound of confidence interval and group indicator for each estimate.