Bug fixes were brought to episensr R package, regarding the use of distributions and computations of odds and risk ratios in probsens.conf function for the probabilistic sensitivity analysis for unmeasured confounder. Improvement on the use of distributions was also brought to other probsens series of functions. Let’s run the example from Modern Epidemiology by Rothman, Greenland & Lash, on page 365-366.
This example is taken from a paper by Greenland et al.
My R package episensr can now also be used through a new Shiny application, episensr_shiny, to more easily assess the effect of biases on epidemiological results.
Not all functions and options are available yet, only selection bias and misclassification of exposure or outcome can be specified. But direct consequences of modifying the bias parameters by moving the sliders can be checked on the 2-by-2 tables and measures of association.
This is therefore still in “beta” but more will come.
A small update for my episensr R package is now available on CRAN. The update focus on misclassification.
First, covariate misclassification is now available, via the function misclassification_cov. For example, the paper by Berry et al. looked if misclassification of the confounding variable in vitro fertilization (IVF), a confounder, resulted wrongly on an association between increase folic acid and having twins. IVF increases the risk of twins, and women undergoing IVF might be more likely to take folic acid supplements, i.