Replace success probabilities by power differences
Source:R/uncertainty_propagation.R
add_power_difference_columns.RdTurns the success probability columns into the difference against the frequentist power at equivalent type I error, with an interval that accounts for the uncertainty of both estimates rather than shifting the success probability interval by the comparator's point estimate.
Arguments
- results_df
A data frame of operating characteristics.
- correlation
Correlation between the two estimates. The default of
0treats them as independent, which is the conservative choice: the two are estimated from simulations sharing a seed, and any positive correlation between them would only narrow the interval.Note separately that the comparator is a deterministic function of a single type I error estimate shared by every row of a panel, so the comparator errors are perfectly correlated across rows whatever this within-row correlation is; points within a panel must not be read as independent.