Alpha oscillations are associated with various cognitive functions. However, the determinants of alpha power variation remain ambiguous, primarily due to its inconsistent associations with autonomic responses and subjective states under different experimental conditions. To thoroughly examine the correlations between alpha power variation and these factors, we implemented a range of experimental conditions, encompassing attentional and emotional tasks, as well as a resting state. In addition to the electroencephalogram data, we gathered a suite of autonomic response measurements and subjective ratings. We employed multivariate linear regression analysis, utilizing autonomic responses and subjective reports as predictors of alpha power. We also subtracted the aperiodic components for better estimation of the power of periodic alpha oscillations. Our results demonstrated that the combined use of autonomic response measurements and subjective ratings effectively predicted the periodic alpha power variation across a range of conditions. These predictions were supported by leave-one-participant-out cross-validation, confirming that multivariate linear relationships can be generalized to new participants. This study demonstrates the links of alpha power variations with autonomic responses and subjective states, suggesting that during investigations of the cognitive functions of alpha oscillations, it is important to consider the potential influences of autonomic responses and subjective states on alpha oscillations.