Repository files navigation

crt2power is an R package that allows users to calculate the statistical power or sample size of their cluster randomized trials (CRTs) with two continuous co-primary outcomes, given a set of input parameters. The motivation for this package is to aid in the design of hybrid 2 studies. Hybrid 2 studies are studies where there are two co-primary outcomes, namely an implementation outcome (such as fidelity or reach) and a health outcome (such as infection rates, or change from baseline health scores). When powering these studies, cluster correlations and the inflation of the Type I error rate must be accounted for.

The five key study design approaches are included in this package that can be used to power hybrid 2 CRTs.

1. P-Value Adjustments for Multiple Testing
2. Combined Outcomes Approach
3. Single 1-Degree of Freedom (DF) Combined Test for Two Outcomes
4. Disjunctive 2-DF Test for Two Outcomes
5. Conjunctive Intersection-Union Test for Two outcomes

For details on the methods listed above, please refer to the publication that discusses these methods by Owen et al., availablehere.

This package is available on CRAN, so it is recommended to run the following code:

install.packages("crt2power")require(crt2power)

If you wish to directly install it from the GitHub repository instead, you can run the following code:

install.packages("devtools")require(devtools)install_github("https://github.com/melodyaowen/crt2power")require(crt2power)

Table of Key Required Input Parameters:

1. When selecting the$\chi^2$-distribution, all methods will use this distribution with the exception of the conjunctive IU test, which will use the multivariate normal (MVN) distribution; when selecting the$F$-distribution, all methods will use this distribution with the exception of the conjunctive IU test, which will use the$t$-distribution.

Each method has a set of functions for calculating the statistical power ($\pi$), required number of clusters per treatment group ($K$), or cluster size ($m$) given a set of input parameters. The names of all functions offered in this package are listed below, organized by study design method.

• calc_pwr_pval_adj() calculates power for this method
• calc_K_pval_adj() calculates number of clusters per treatment group for this method
• calc_m_pval_adj() calculates cluster size for this method

• calc_pwr_comb_outcome() calculates power for this method
• calc_K_comb_outcome() calculates number of clusters per treatment group for this method
• calc_m_comb_outcome() calculates cluster size for this method

• calc_pwr_single_1dftest() calculates power for this method
• calc_K_single_1dftest() calculates number of clusters per treatment group for this method
• calc_m_single_1dftest() calculates cluster size for this method

• calc_pwr_disj_2dftest() calculates power for this method
• calc_K_disj_2dftest() calculates number of clusters per treatment group for this method
• calc_m_disj_2dftest() calculates cluster size for this method

• calc_pwr_conj_test() calculates power for this method
• calc_K_conj_test() calculates number of clusters per treatment group for this method
• calc_m_conj_test() calculates cluster size for this method

• run_crt2_design(output = "power", ...) calculates power for all 5 methods
• run_crt2_design(output = "K", ...) calculates number of clusters per treatment group for all 5 methods
• run_crt2_design(output = "m",...) calculates cluster size for all 5 methods

# Example of using the combined outcomes approach for calculating powercalc_pwr_comb_outcome(dist = "Chi2", K = 8, m = 50, alpha = 0.05,                      beta1 = 0.2, beta2 = 0.4, varY1 = 0.5, varY2 = 1,                      rho01 = 0.05, rho02 = 0.1, rho1 = 0.01, rho2 = 0.1,                       r = 1)# Example of using the single weighted 1-DF test for calculating Kcalc_K_single_1dftest(dist = "F", power = 0.9, m = 70, alpha = 0.05,                      beta1 = 0.4, beta2 = 0.3, varY1 = 1.5, varY2 = 0.5,                      rho01 = 0.1, rho02 = 0.07, rho1 = 0.05, rho2  = 0.3,                       r = 2)# Example of using conjunctive IU test for m calculationcalc_m_conj_test(dist = "MVN", power = 0.8, K = 10, alpha = 0.05,                  beta1 = 0.4, beta2 = 0.4, varY1 = 0.5, varY2 = 1,                  rho01 = 0.05, rho02 = 0.1, rho1 = 0.07, rho2  = 0.9,                  r = 1, two_sided = TRUE)# Example of calculating power based on all five methodsrun_crt2_design(output = "power", K = 6, m = 70, alpha = 0.05,                 beta1 = 0.4, beta2 = 0.4, varY1 = 0.5, varY2 = 0.5,                 rho01 = 0.1, rho02 = 0.1, rho1 = 0.07, rho2 = 0.9, r = 1)

For questions or comments, please email Melody Owen atmelody.owen@yale.edu, or submit an issue to this repository.