This package calculates an estimate of the relative risk based on the ratio of two median unbiased estiamtes for a proportion. In addition, we enumerate the exact bootstrap confidence interval for estimate. Details for the methods are available in Carter et al (2010). This package replaces the previous SAS MACRO supplied with the paper.
The manuscript shows example data of from a renal transplant study. The study was monitoring the events of severe hypoglycemia at regular DSMB meetings. There was a desire to summarize the risk of hypoglycemia between the intensively managed patients relative to the routine glucose management strategy with the relative risk. At the first time period, there were 0/3 and 0/4 events in these two groups, respectively. Despite the zero cells, the MUE-based estimate of relative risk is well defined. In particular
library(mueRelativeRisk)
mue_rr(y1=0, n1=3, y2=0, n2=4, alpha=0.05)
#> prop1 p1_mue prop2 p2_mue mue_rr_estimate mue_ci_lower
#> 1 0/3 0.1031497 0/4 0.07955179 1.296636 0.2067111
#> mue_ci_upper
#> 1 8.055469The function does not currently extend to a vector of values for y1, y2, n1, and n2. However, such operations can be programmed outside the function using various combinations of apply.
Future releases of the package will include the MUE-based estimate for the odds ratio and risk difference.