Impeachement Polls during Trump's First Term

This report covers: Part A: daily ATM cash forecasting for May 2010. Part B: monthly residential power forecasting for 2014. Bonus Part C: hourly waterflow aggregation, stationarity assessment, and one-week-ahead forecasting. Across all three parts, the modeling work depended on fixing data-format and structure problems before any forecasts could be trusted. The ATM file required Excel date conversion and removal of trailing non-series rows, the power file required interpolation of a missing monthly usage value, and the waterflow files required conversion from Excel serial timestamps into regular hourly series. Those cleanup steps were not cosmetic. Each one addressed a specific failure mode that would otherwise distort the time index, break the tsibble structure, or cause the forecasting functions to evaluate the wrong objects.