clintools (version 0.9.10.1)

rrGcomp: Relative risk derived by G-computation (rrGcomp)

Description

rrGcomp() is a small function which generates population-level (marginal) relative risks derived by G-computation. For models with random effects mixed-effects generalized linear model with a logit link with adjustment for stratification variables will be used, while those without random effects a logistic regression will be used. The code is based on the method used in the paper by Dankiewicz et al. (2021) N Engl J Med. Jun 17;384(24):2283-2294. (PubMed

Usage

rrGcomp(df, outcome_col, group_col,
fixed_strata = NULL, random_strata = NULL,
nbrIter = 5000, conf_level = 0.95)

Value

Returns a list with relative risk (rr), simulated rr (simRR), lower- and upper confidence level (simLCL/simUCL), and the p-value (p_val)

Arguments

df

the individual participant dataframe

outcome_col

column name for the outcome column

group_col

column name for the group column

fixed_strata

list of column names for the fixed effect stratification columns

random_strata

list of column names for the random effect stratification columns

nbrIter

number of iterations to be used in the G-computation. The original paper used 5000, which is also the default.

conf_level

the confidence level to be reported.

Examples

df <- sRCT(n_sites=3,n_pop=50)
rrGcomp(df,outcome_col="outcome",group_col="Var1",random_strata="site",nbrIter=10)