epijim/episheet

R functions based on episheet and the text book Modern Epidemiology, Second Edition, by Kenneth J. Rothman and Sander Greenland.

Hi there 👋

My 🌐 is at epijim.uk

I run a cron job that makes commits - so my streak is not human in nature... (I should probably make a bot account)

episheet's Issues

Consider adding function(s) to do standardisation

Problem statement

Ken Rothman's episheet has a tab for calculating standardised rates and proportions. This package has no such functions.

Proposed solution

Add one or more functions to calculate standardised rates and proportions.
See Ch. 15 from p265 of 2nd Ed Modern Epidemiology.
It should fit with tidyverse so that can do grouped calculations.
Table 15-2, p268 has sample data (which I think already exists in package?), same as on standardisation tab in Excel episheet.

Redundant ci_level argument in risk function.

Using example data in package help

dat <- data.frame(
    exposure_var = c(rep(1, 8), rep(0, 5), rep(1, 98), rep(0, 115)),
  outcome_var = c(rep(1, 8), rep(1, 5), rep(0, 98), rep(0, 115)),
  stringsAsFactors = FALSE
) 

Output with default ci_level

> risk(data = dat, exposure = exposure_var, outcome = outcome_var)

## # A tibble: 2 x 10
##  exposure_var     n outcome  risk risk_ratio rr_lci rr_uci risk_diff rd_lci rd_uci
##         <dbl> <int>   <dbl> <dbl>      <dbl>  <dbl>  <dbl>     <dbl>  <dbl>  <dbl>
## 1            0   120       5  4.17       1     0.337   2.96      0     -6.20   6.20
## 2            1   106       8  7.55       1.81  0.611   5.37      3.38  -2.82   9.58

Output with ci_level set to 99

> risk(data = dat, exposure = exposure_var, outcome = outcome_var,ci_level = 99)
## # A tibble: 2 x 10
##   exposure_var     n outcome  risk risk_ratio rr_lci rr_uci risk_diff rd_lci rd_uci
##          <dbl> <int>   <dbl> <dbl>      <dbl>  <dbl>  <dbl>     <dbl>  <dbl>  <dbl>
## 1            0   120       5  4.17       1     0.337   2.96      0     -6.20   6.20
## 2            1   106       8  7.55       1.81  0.611   5.37      3.38  -2.82   9.58

I believe the line that reads z <- abs(qnorm(((100 - 95)/2)/100)) in the function should be z <- abs(qnorm(((100 - ci_level)/2)/100)). Editing the function as such produces different confidence limits when ci_level is set.