Do the above factors effects combine in a linear way to predict EABR latency?

A multiple linear regression analysis using subjects’ actual data (instead of groups) was completed for the factors that were found to have a significant effect on EABR latency with the analysis of variance. This investigated whether EABR latency may be predicted from a linear combination of those factors. For analysis on wave IIIe, the normality test and the constant variance test failed with respectively p = 0.008 and p < 0.001. Both tests passed for analysis on wave Ve (with both p > 0.05) and interval IIIe-Ve (with both p > 0.05).

Latency of wave IIIe can be predicted from the following linear combination, which explains 37 % of the variance (adjusted R² = 0.37): 2.14 + 0.028 x stimulation site + 0.016 x auditory deprivation + 0.099 x audiogram type + 0.00072 x M level. The contribution to this combination was significant for electrode site (p < 0.001) and the duration of deprivation (p < 0.001), but not for audiometric data (p = 0.079) or M level (p = 0.06). For latency of wave Ve, latency of wave IIIe was added to the analysis because of the striking resemblance between the effects of the various factors on both waves’ latencies. Parameters having no significant effect on interval IIIe-Ve latency were removed. Wave Ve latency was found to be predictable from the following linear combination, which explains 56 % of the variance (adjusted R² = 0.56): 2.09 + 0.94 x wave IIIe latency + 0.00076 x M level + 0.041 x audiogram type. Contribution to this combination was significant for wave IIIe latency (p < 0.001) but not for audiometric data (p > 0.05) and M level (p > 0.1). Interval IIIe-Ve cannot be predicted from a linear combination since the equation found involved no significant contribution from M level (p = NS) nor audiometric data (p = NS) and did not explain any variance (adjusted R² = 0).