<Fig. 2. here>
Figure 2. Contrast sensitivity results for controls
(green) and patients (red) for longer presentation times (500 ms) .
Results were qualitatively the same as with the short presentation data
(33 ms), suggesting, again, that spatial frequency interactions depend
on whether the data were log transformed and whether groups differed in
visual acuity.
To show that these results generalize to other viewing conditions, we
also considered data from the longer presentation condition (500 ms).
The interactions and main effects were similar to before (see Figure 2;
Tables S3). In particular, using all subjects, the interaction was
significant in opposite ways depending on whether the log-transform was
applied or not (raw data: F (1,141)=7.6, p =.006,\(\eta_{p}^{2}\) =.05; log-transformed: F (1,141)=17.7,p <.001, \(\eta_{p}^{2}\)=.11). Using the matched
subsample of controls, we found that the interaction showed reduced
contrast sensitivity deficits at higher spatial frequencies before but
not after the log-transform (raw data: F (1,100)=4.7,p =.03, \(\eta_{p}^{2}\)=.05; log-transformed:F (1,100)=2.6, p =.11, \(\eta_{p}^{2}\)=.03). In each of the
above cases, there were main effects of group and spatial frequency (allp <.001; all \(\eta_{p}^{2}\)>.14). GEEs
yielded results that were qualitatively the same as the log-transformed
results with one exception: using the subsample matched on acuity, there
was now an interaction with spatial frequency such that patient deficits
tended to worsen at the higher spatial frequency (B =-.436,SEB =.19, Wald Chi-square(1)=5.26, p =.02).
However, adding logMAR as a covariate would nudge the interaction out of
significance territory (B =-.37, SEB =.20,
Wald Chi-square(1)=3.7, p =.06). Moreover, if we were to more
closely match groups on acuity, then the interaction would likely be
pushed further from significance by reducing the high spatial frequency
group difference. To summarize, the results for the long presentation
condition were largely the same as the short presentation condition.
Our focus in the study was on spatial frequency interactions. However,
it is worth noting that–in all the foregoing analyses, including acuity
matching–patients had worse contrast sensitivity than controls (allp <.05). If we were to remove patients whose vision was
worse than 20/25–so that the groups would become even more closely
matched on logMAR acuity (MHC=.02,
MSZ=.02, p =.96, Hedges’ g=.01; n =34 SZ, 58
HC)–the log-transformed ANOVA results would again reveal a strong
group deficit in the short duration condition
(\(\eta_{p}^{2}\)>.10, p <.004) and the
long duration condition (\(\eta_{p}^{2}\)>.11,p <.004) with or without a logMAR covariate.