Lastly, two previously effects that were only marginally significant under parametric testing reached significance under non-parametric testing: the median direction noise (°) was lower (8.4%, W (16) = −75, p = 0.0389) and median direction asymmetry higher (58.3%, W (16) = 76, p = 0.0485) with binocular viewing.
For the pupillary time constants, there were a few missing values which reduced the power of the coin-flip testing, however for the contraction τ , there were 10 out of 15 points on one side of the diagonal ( p = 0.0592) and for the dilation τ, 10 out of 13 points ( p = 0.0461), indicating nearly significant and significant effects, respectively.
Their post-hoc t-testing found that their observed U-D asymmetry (downward preference) for latency was not quite significant (ours was), for steady-state gain was significant at higher speeds (their Table 2) (ours approached significance), and for smoothness was significant (ours was not).