A statistical study of the Individual Household Electric Power Consumption dataset from the UCI Machine Learning Repository: four years of minute-by-minute electrical measurements (active and reactive power, voltage, current) from one household.
The analysis runs in two phases. The first is classical statistics: sampling from the population, testing hypotheses about consumption, comparing variances, and fitting regressions to find linear relationships between the electrical variables. The second brings in multivariate techniques: principal component analysis to cut the dimensionality of the measurements, clustering to find natural groups in consumption, and ANOVA to test whether those groups really differ.
Done entirely in R, with written reports on the method and findings of each phase.