interactionRCS facilitates interpretation and
presentation of results from a regression model (linear, logistic, Cox)
where an interaction between the main predictor of interest X (binary or
continuous) and another continuous covariate Z has been specified. In
particular, interactionRCS allows for basic interaction
assessment (i.e. log-linear/linear interaction models where a product
term between the two predictors is included) as well as settings where
the second covariate is flexibly modeled with restricted cubic splines.
Confidence intervals for the predicted effect measures (beta, OR, HR)
can be calculated with either bootstrap or the delta method. Lastly,
interactionRCS produces a plot of the effect measure over
levels of the other covariate.
To install the latest version of interactionRCS, type
the following lines in a web-aware R environment.
if(!"devtools" %in% rownames(installed.packages())){
install.packages("devtools")
}
devtools::install_github("https://github.com/gmelloni/interactionRCS.git")
library(interactionRCS)After estimating a regression model (linear, logistic, Cox) such as
model<-glm(y~ ...) estimate and plot interactions
with:
int<-estINT(model=model, ...)
plotINT(int, ...)For a detailed introduction to interactionRCS and code
examples please refer to this vignette
Giorgio Melloni, Andrea Bellavia
TIMI study group, Department of Cardiovascular Medicine, Brigham and Womens Hospital / Harvard Medical School