Prescribed fire is increasingly used as a tool for managing and restoring native plant species and controlling invasive exotics in grasslands worldwide. Despite growing interest in this practice, assessing the impacts of such large-scale disturbances is problematic. First, it is often difficult to collect randomized, well-replicated data from single, large-scale prescribed burns, which can lead to data being dismissed as un-analyzable due to pseudoreplication or analyzed using inappropriate statistical models.  Second, well-designed experiments involving fire are often difficult to implement due to logistical and financial constraints. Here, we assess the effects of fire in northern California grasslands using a large-scale comparative study and a smaller-scale experiment designed to look at the short-term effects of fire. We also use spatial statistical methods to estimate and correct for spatial autocorrelation (SAC) that exists in our data. Overall, we found that the biases due to SAC were low in the comparative study, even with the lack of appropriate replication, and that our experiment was no more efficient at reducing SAC than the comparative approach. Thus, even in the absence of well-replicated experiments, statistical power can be successfully increased for monitoring studies understanding and accounting for SAC.  We also found that both approaches provided similar results: prescribed fire was effective at reducing a dominant invasive grass (Taeniatherum caput-medusae) and increasing native annual grasses and forbs, but also increased exotic annual forbs. In summary, our results demonstrate that prescribed fire may be an effective tool for meeting restoration and management objectives for grasslands, yet may also have unintended consequences such as enhancing invasion by non-native species.