ObjectiveTo examine how known causal factors (exercise self-efficacy, balance, walking capacity) affect outcomes (moderate intensity physical activity, community access) in people with Parkinson's disease (PD): through a direct pathway, indirectly through potential mediators (non-motor impairments), or through combined direct and mediated paths. DesignCausal mediation analyses using baseline and three-month data from pooled treatment groups in a randomized controlled trial. SettingData were collected at two university clinical research centers. ParticipantsOne hundred thirty-eight people with PD. InterventionsHome and community-based walking and strength exercise program. Main Outcome Measure(s)Exposures were exercise self-efficacy, walking ability, and balance. Potential mediators included pain, fatigue, mood (anxiety, depression, affect, apathy), stigma, and cognition (executive function and cognitive flexibility). Outcomes were physical activity and community access behaviour at three-month follow-up. Separate models were developed for each causal factor-mediator-outcome combination. To minimise bias, all models were adjusted for known confounders (comorbidities, freezing of gait, severity of motor impairments, and/or age) and baseline values of the outcomes. ResultsSelf-efficacy of walking duration had a causal relationship with moderate intensity physical activity through direct and combined paths (p<.001). Walking capacity had a causal relationship with community access through direct (p=.03-.04) and combined (p=.02-.03) paths. Balance did not affect community access (p>.05). There were no significant mediation effects through indirect pathways for either outcome. ConclusionsThe effect of known causal factors on physical activity and community access was not mediated by non-motor impairments. Walking self-efficacy and walking capacity remain the primary intervention targets for improving physical activity and community access, respectively, in people with PD.