Abstract: This article shows how recent advances in statistical analytical
strategies could be applied to correlational or observational data collected
from non-experimental designs in order to provide convergent validity for
causal inferences regarding "change" in two broad contexts. The first context
refers to modeling causal relationships between constructs, specifically on
relationships that go beyond the "bivariate prediction paradigm". In this
context, mediation analyses, interaction analyses, combination of interactions
and mediations, and structural equation modeling were discussed. The second
context refers to modeling the causes of changes over time. In this context,
fundamental questions on changes over time were explicated, limitations of
traditional techniques for analyzing changes over time were illustrated, and
latent variable approaches to modeling changes over time were discussed.