The nonaccommodated CLS decreased significantly with age in both the nasal (−0.012 ± 0.004 mm/y;
P < 0.03) and temporal (−0.016 ± 0.004 mm/y;
P < 0.01) quadrants and tended to do so at a slightly faster rate in the temporal quadrant
(Figs. 5A 5B) . The nonaccommodated nasal and temporal CLS was significantly correlated with the accommodative amplitude of each monkey
(Fig. 6A) ; the greater the CLS the greater the accommodative amplitude. Similar results were seen in the maximally
(Fig. 6B)and supramaximally
(Fig. 6C)stimulated states. Because CLS and accommodative amplitude covaried with age, a multiple regression analysis that models accommodation as a linear function of age and nonaccommodated CLS was undertaken. A mixed model (using SAS Proc Mixed; SAS, Cary, NC) that recognizes that measurements taken from two eyes of the same monkey may correlate, was used, with the thought that the decrease in nonaccommodated CLS with age might explain the decrease in accommodative amplitude over and above what age could do alone. The multiple regression coefficient of nonaccommodated temporal (but not nasal) CLS was significantly different from 0.0 (
P < 0.03), indicating that age and nonaccommodated temporal CLS together are better predictors of accommodative amplitude than is age alone (
Table S1). A similar multiple regression analysis using CLS measured during maximal stimulation showed significance in both quadrants, indicating that age and CLS during maximal stimulation together are better predictors of accommodative amplitude than is age alone (Table S1). Similar results were seen at the supramaximal stimulus level. The difference between the CLS in the nonaccommodated state and during maximal and supramaximal stimulation did not decline significantly with age in either quadrant. However, in the nasal quadrant, the supramaximally accommodated CLS minus the maximally accommodated CLS tended to decline with age (
P = 0.053). The results of the multiple regression analysis and the corresponding F-statistic show that age (which typically has much larger F-statistics than CLS) explains nearly all of the variation in accommodative amplitude. However, a stated earlier, CLS was important in predicting accommodative amplitude over and above what age could do alone.
The model diagnostics were performed and, based on the residual plots, the proposed models are not unduly affected by variance fluctuations.
28 Thus, these models would be reasonable for the underlying data. Because these regression models take the relatedness of the eyes into account, it is not possible to divide the total model variance into components based on individual independent variables. However, the more significant the probability for the individual independent variable, the more it would contribute to the explanation of the variance.