<Fig. 1. here>
Figure 1. Contrast sensitivity results for controls (green) and patients (red) for briefly presented gratings (33 ms) . (A, B) When using non-transformed data, patients exhibited worse contrast sensitivity deficits at low versus high spatial frequencies. (B) Such an interaction also emerged when using a subset of controls (n=34) that were approximately matched to patients on visual acuity; (C, D) When the data were log-transformed, the sign of the interaction reversed. However, this interaction disappeared when using the acuity-matched set of controls.
A critical caveat is that patients’ acuity differed from controls by about one line on an eye chart (t (141)=7.1,p <.001, Hedges’ g =1.2, logMAR difference =.107). To consider whether acuity differences could explain the spatial frequency interaction, we simply removed healthy controls whose acuity was better than 20/20 (logMAR<0) so that the remaining controls (n =34) were approximately matched to patients on this variable (t (99.5)=1.9, p =.06, Hedges’ g=.34, logMAR difference =.03). An ANOVA on the logged data revealed main effects of spatial frequency and subject group, but no interaction (spatial frequency: F (1,100)=1467, p <.001;\(\eta_{p}^{2}\) =.94; group: F (1,100)=12.9,p <.001, \(\eta_{p}^{2}\) =.11; interaction:F (1,100)<0.01, p =.99, \(\eta_{p}^{2}\)<.00). The GEE also revealed main effects of spatial frequency and group (B =-2.91, SEB =.22, Wald Chi-Square(1)=177.15, p <.001; B =-.26,SEB =.12, Wald Chi-Square(1)=4.48,p <.05), with no interaction (B =-.025,SEB =.24, Wald Chi-square(1)=.01,p >.9). There was also no interaction when logMAR acuity was added as a covariate to the ANOVA (spatial frequency:F (1,99)=108.32, p <.001, \(\eta_{p}^{2}\)=.522; group: F (1,99)=11.00, p =.001, \(\eta_{p}^{2}\)=.100; interaction: F (1,99)=2.69, p= .10, \(\eta_{p}^{2}\)=.026) or to the GEE (B =-.059, SEB =.26, Wald Chi-square(1)=.05, p >.8).