3.5.2 Regression Analysis
To further investigate the relationship between adaptation and MMN, a backward stepwise regression was performed to clarify which and how each adaptation effect contributes to the MMN. First, the six adaptation variables measured by amplitudes were entered into the regression model, while the MMN peak amplitude acted as the DV. Only N1 initial adaptation (β (standardized beta) = 0.69; p < .05) and P2 subsequent adaptation amplitudes (β = 0.26; p < .10) survived in the final model (F (2,26) = 11.03; p< .001). According to the squared semipartial correlation, these two variables explained more than 50% of the MMN variance (44.89% and 6.55% respectively). The standardized coefficients of each variable and the adjusted R squares of the initial and final models of the amplitudes of adaptation and MMN are shown in Table 3a.
Likewise, a backward stepwise regression was performed to examine which and how each adaptation latency contributes to the MMN amplitude by entering the six variables related to latency as IVs. Only the subsequent adaptation latency of P2 (β = 0.33; p< .10) stayed in the final model (F (1,27) = 3.40;p = .08). The semipartial correlation bespoke that it uniquely explained 11.16% variance of the MMN peak amplitude. Table 3b lists the standardized coefficients of each variable and the adjusted R squares of the initial and final models of the adaptation latency and MMN amplitude. The scatter plots illustrating the relationships between the MMN amplitude and the significant variables in the correlation and regression, including the amplitudes of N1 initial adaptation, P1 initial adaptation, P2 subsequent adaptation, and the latency of P2 subsequent adaptation were shown in Figure 4.