Precipitation is influenced by multiple large-scale natural processes. Many of these large-scale precipitation ''drivers'' are not independent of one another, which complicates attribution. Moreover, it is unclear whether natural interannual drivers alone can explain the observed longer-term precipitation trends or account for projected precipitation changes with global warming seen in climate models. Separating the main interannual drivers from processes that may prevail on longer time scales, such as a poleward circulation shift or increased specific humidity, is essential for an improved understanding of precipitation variability and for making longer-term predictions.In this study, an objective approach to disentangle multiple sources of large-scale variability is applied to Australian precipitation. This approach uses a multivariate linear independence model, involving multiple linear regressions to produce a partial correlation matrix, which directly links variables using significance thresholds to avoid overfitting. This is applied to regional winter precipitation in Australia as a test case, using the ECMWF Interim Re-Analysis (ERA-Interim) and Australian Water Availability Project datasets. Traditional drivers and several drivers associated with the width of the tropics are assessed.The results show that the web of interactions implied by correlations can be simplified using this multivariate linear independence model approach: the total number of apparent precipitation drivers was reduced in each region studied, compared to correlations meeting the same statistical significance. Results show that the edge of the tropics directly influences regional precipitation in Australia and also has an indirect influence, through the interaction of the subtropical ridge and atmospheric blocking. These results provide observational evidence that changes associated with an expansion of the tropics reduce precipitation in subtropical Australia.