Longitudinal count data are often collected in a variety of health domains. This repository contains code to estimate sample size needed to compare dynamic treatment regimens using longitudinal count outcomes from a Sequential Multiple Assignment Randomized Trial (SMART). A particular focus of this repository is on longitudinal count data having overdispersion.

A pair of dynamic treatment regimens embedded in a planned SMART (aka. 'EDTRs') can be compared using differences in end-of-study means, or more generally, differences in a weighted average of means across various time points, which we denote as ; Q is simply shorthand for 'quantity', e.g., denotes the quantity difference in end-of-study means.

CountSMART is about a Monte Carlo simulation-based approach developed to estimate sample size required to attain power of to the test of the null against the alternative at type-I error rate .

This repository contains code implementing CountSMART methodology and simulation studies examining the validity of the approach.

1. The collection of packages and their version numbers used for this repository are recorded in the renv.lock file. The package, renv, can facilitate installation of these packages in the machine of end-users of this repository. See renv package documentation here for more details: https://rstudio.github.io/renv/articles/renv.html

1. Create a new R file named 'paths.R' and save this file within the root directory of the repository (usually where the .Rproj file is located).
2. Within 'paths.R', set the value of the following variables below by replacing the three dots '...' with the appropriate directory.

• path.output_data = ".../output"
• path.code = ".../code"
• path.plots = ".../plots"

Note that 'paths.R' is included in the '.gitignore' file, preventing any user-specific directories from being displayed in the repository. Also, since 'paths.R' is included in the '.gitignore' file, a new 'paths.R' file would need to be created by each end-user of the repository.

File Name	Brief Description
input-utils.R	Contains a function for checking validity of time-specific means and proportion of zeros provided as inputs to the sample size estimation procedure.
datagen-utils.R	Collection of functions to generate potential outcomes and observed outcomes.
analysis-utils.R	Collection of functions to 'analyze' data from a SMART.

File Name	Brief Description
calc-covmat.R	Calculate estimated covariance matrix.
calc-corr-params-curve.R	Implement simulation to estimate relationship between and and the relationship between and .
calc-truth-beta.R	Calculate true value of parameters in a model for the mean trajectory of dynamic treatment regimens embedded in a SMART, implied by inputs provided to Monte Carlo simulation.
calc-truth-contrasts.R	Calculate true value of in a model for the mean trajectory of dynamic treatment regimens embedded in a SMART, implied by inputs provided to Monte Carlo simulation.
plot-truth-deltaQ.R	Wrapper for calc-truth-beta.R and calc-truth-contrasts.R. Visualize true mean trajectory of each dynamic treatment regimen embedded in a SMART, implied by inputs provided to Monte Carlo simulation.
geemMod.R	Modification of the geem.R script from the R package geeM: setting the additional argument fullmat=TRUE allows custom specification of working correlation matrix for each participant-time.

File Name	Brief Description
create-scenarios-ar.R	A script to create simulation study scenarios.
calculate-dispersion-param.R	A script to calculate the value of the negative binomial dispersion parameter in the different simulation scenarios.
simulation-study-pipeline-ar.R	A script to document and run steps in the simulation study pipeline.
sim_size_test	A directory containing a collection of scripts to execute simulation studies concerning empirical type-I error rate. Results of simulation studies are also provided here (e.g., power.csv file).
sim_vary_effect	A directory containing a collection of scripts to execute simulation studies investigating how power changes as specific choices of are increased across a grid of total sample sizes N=100, 150, 200, ..., 550. Results of simulation studies are also provided here (e.g., power.csv file).
sim_vary_n4	A directory containing a collection of scripts to execute simulation studies investigating whether power is sensitive to a violation in our working assumption on the number of individuals who would not respond to either first-stage intervention option. Results of simulation studies are also provided here (e.g., power.csv file).
sim_vary_eta	A directory containing a collection of scripts to execute simulation studies investigating whether power is sensitive to the actual value of given fixed value of and N. Results of simulation studies are also provided here (e.g., power.csv file).

File Name	Brief Description
create-scenarios-exch.R	A script to create simulation study scenarios.
calculate-dispersion-param.R	A script to calculate the value of the negative binomial dispersion parameter in the different simulation scenarios.
simulation-study-pipeline-exch.R	A script to document and run steps in the simulation study pipeline.
sim_vary_effect	A directory containing a collection of scripts to execute simulation studies investigating how power changes as specific choices of are increased across a grid of total sample sizes N=100, 150, 200, ..., 550. Results of simulation studies are also provided here (e.g., power.csv file).

File Name	Brief Description
data-viz-pipeline-ar.R	A script to document and run steps in the data visualization pipeline.
plot-sim-size-test.R	A script to plot results in sim_size_test
plot-sim-vary-effect.R	A script to plot results in sim_vary_effect
plot-sim-vary-n4.R	A script to plot results in sim_vary_n4
plot-sim-vary-eta.R	A script to plot results in sim_vary_eta
corviz_sim_size_test	A directory containing visualization of empirical correlation matrices. Values of parameters identical to those used to obtain results in the directory sim_size_test were used to calculate the values displayed, except that N was fixed to 1000.
corviz_sim_vary_effect	A directory containing visualization of empirical correlation matrices corresponding to each scenario considered. These results accompany those in the directory sim_vary_effect. Values of parameters identical to those used to obtain results in the directory sim_vary_effect were used to calculate the values displayed, except that N was fixed to 1000.
corviz_sim_vary_eta	A directory containing visualization of empirical correlation matrices corresponding to each scenario considered. Values of parameters identical to those used to obtain results in the directory sim_vary_eta were used to calculate the values displayed, except that N was fixed to 1000.

File Name	Brief Description
data-viz-pipeline-exch.R	A script to document and run steps in the data visualization pipeline.
plot-sim-vary-effect.R	A script to plot results in sim_vary_effect