Use a scatterplot and a suitable correlation method to describe an association, while recognizing that correlation does not prove causation.
Use one-way ANOVA to compare group means when the design and model assumptions fit your data, then interpret the result with diagnostics and context.