test/testing notes 2020-07-08.md

In UCL/mimix: Reference-Based Imputation for Longitudinal Clinical Trials with Protocol Deviation

Ian's mimix testing notes, 8jul2020

updated to v0.0.7 - had to delete package ellipsis at N:\R\win-library\3.6.then reinstall it, in order for install_github() to work.

help file

• see word doc

front page

causal

• *** need a timescale option: is k1 per unit time or per visit?

• i compared k0=k1=0 with J2R: perfect agreement

• *** i compared k0=k1=1 with CIR: disagreement after discontinuation when there were intermittent missing data before discontinuation.

• *** I was unclear how to verify the results more broadly. Do you have code for the examples in the paper?

delta

• *** setting dlag of the wrong length leads to missing imputed values and "WARNING!!! unimputed data values, possibly due to mis-specified delta". Should catch this error and fail, as for delta of wrong length.

• tested for arbitrary values of delta and dlag and verified that the increases in imputed values are as expected after discontinuation and are zero before discontinuation

main package

• 3 arms: changed first 200 obs in asthma data to a 3rd arm. Verified agreement with Stata package to within MC error with M=100.

• Also noted R package is MUCH faster than Stata.

• added second covariate (=base^2). Verified agreement with Stata package to within MC error with M=100.

• multiplied outcome by 1000. Results are exactly multipled by 1000.

• after conversion to mids format, attributes such as pmm are wrong. However this is NOT A PROBLEM as it is happening outside our function.

to discuss

• allow incomplete covariates? or add a recommendation to use mean imputation or missing indicator?

UCL/mimix documentation built on Sept. 7, 2024, 5:26 p.m.