The basic strategy of the Dunedin Study involves testing for causal relations within this correlational design. We adopt a stepwise approach. First we document that a basic association between two constructs exists, striving for gold-standard (for the time) measurement. Second, we document temporal sequence, as to whether the putative causal variable precedes the outcome variable. Third, we look for a dose–response contingency between the putative cause and outcome. Fourth, we attempt to rule out as many rival causal explanations as possible by introducing control variables into the analyses from our extensive database. Fifth, we test whether putative causal experiences are associated with intra-individual change. Sixth, we try to establish specificity by substituting alternative dependent, then independent variables. Finally, we posit a plausible explanatory process and test this using mediation analyses. By following this general approach we hope to maximise the robustness of our research findings. Ultimately, of course, how ‘true’ a finding is will be determined via independent replication. However, this general approach (augmented by multiple checks including sensitivity analyses) increases the likelihood of successful replication.
