Benefit direction implied by the null space
Source:R/npp_kl_calibration.R
benefit_sign_from_null_space.RdA case study records which direction of the treatment effect is
beneficial only through null_space: "left" means the null hypothesis is
the half line below theta_0, so the trial succeeds when the effect is
large, and "right" is its mirror. The maximum tolerable discrepancy has to
be taken on the side that moves the target towards the null, so it needs that
direction as a sign. The same mapping is written as an alternative hypothesis
by alternative_from_null_space().