tunevt {tehtuner}

Fit a tuned Virtual Twins model

Description

tunevt fits a Virtual Twins model to estimate factors and subgroups associated with differential treatment effects while controlling the Type I error rate of falsely detecting at least one heterogeneous effect when the treatment effect is uniform across the study population.

Usage

tunevt(
  data,
  Y = "Y",
  Trt = "Trt",
  step1 = "randomforest",
  step2 = "rtree",
  alpha0,
  p_reps,
  threshold = NA,
  keepz = FALSE,
  parallel = FALSE,
  ...
)

Arguments

data

a data frame containing a response, binary treatment indicators, and covariates.

Y

a string specifying the name of the column of data contains the response.

Trt

a string specifying the name of the column of data contains the treatment indicators.

step1

character strings specifying the Step 1 model. Supports either "lasso", "mars", "randomforest", or "superlearner".

step2

a character string specifying the Step 2 model. Supports "lasso", "rtree", "classtree", or "ctree".

alpha0

the nominal Type I error rate.

p_reps

the number of permutations to run.

threshold

for "step2 = 'classtree'" only. The value against which to test if the estimated individual treatment effect from Step 1 is higher (TRUE) or lower (FALSE).

keepz

logical. Should the estimated CATE from Step 1 be returned?

parallel

Should the loop over replications be parallelized? If FALSE, then no, if TRUE, then yes. Note that running in parallel requires a parallel backend that must be registered before performing the computation. See the foreach documentation for more details.

...

additional arguments to the Step 1 model call.

Details

Virtual Twins is a two-step approach to detecting differential treatment effects. Subjects' conditional average treatment effects (CATEs) are first estimated in Step 1 using a flexible model. Then, a simple and interpretable model is fit in Step 2 to model either (1) the expected value of these estimated CATEs if step2 is equal to "lasso", "rtree", or "ctree" or (2) the probability that the CATE is greater than a specified threshold if step2 is equal to "classtree".

The Step 2 model is dependent on some tuning parameter. This parameter is selected to control the Type I error rate by permuting the data under the null hypothesis of a constant treatment effect and identifying the minimal null penalty parameter (MNPP), which is the smallest penalty parameter that yields a Step 2 model with no covariate effects. The 1-alpha0 quantile of the distribution of is then used to fit the Step 2 model on the original data.

Value

An object of class "tunevt".

An object of class "tunevt" is a list containing at least the following components:

call

the matched call

vtmod

the model estimated by the given step2 procedure fit with the permuted tuning parameter for the estimated CATEs from the step1 model. See vt2_lasso, vt2_rtree, or vt2_ctree for specifics.

mnpp

the MNPP for the estimated CATEs from Step 1.

theta_null

a vector of the MNPPs from each permutation under the null hypothesis.

pvalue

the probability of observing a MNPP as or more extreme as the observed MNPP under the null hypothesis of no effect heterogeneity.

z

if keepz = TRUE, the estimated CATEs from the step1 model.

References

Foster JC, Taylor JM, Ruberg SJ (2011). “Subgroup identification from randomized clinical trial data.” Statistics in Medicine, 30(24), 2867–2880. ISSN 02776715, doi:10.1002/sim.4322.

Wolf JM, Koopmeiners JS, Vock DM (2022). “A permutation procedure to detect heterogeneous treatment effects in randomized clinical trials while controlling the type I error rate.” Clinical Trials, 19(5), 512-521. ISSN 1740-7745, doi:10.1177/17407745221095855, Publisher: SAGE Publications.

Deng C, Wolf JM, Vock DM, Carroll DM, Hatsukami DK, Leng N, Koopmeiners JS (2023). “Practical guidance on modeling choices for the virtual twins method.” Journal of Biopharmaceutical Statistics. doi:10.1080/10543406.2023.2170404.

Examples

data(tehtuner_example)
# Low p_reps for example use only
tunevt(
  tehtuner_example, step1 = "lasso", step2 = "rtree",
  alpha0 = 0.2, p_reps = 5
)