In our study we applied the split sample method to the Wilson interval for a binomial proportion. This interval has
excellent performance, with competitive coverage and length properties, and is often recommended for general use (Brownet al., 2001; Newcombe, 2012). We note however that it can be more or less uniformly improved upon by using a leveladjusted
Clopper–Pearson interval (Thulin, 2014a) instead; we opted to use the Wilson interval in our study simply because
it is less computationally complex.