Essays on Time Series Analysis

Abstract

I contribute to the literature on time series analysis in Political Science in this dissertation. My research broadly focuses on three things: improving inferences from stationarity testing, implementing an all-encompassing model that accounts for common problems faced by time series analysts, and assessing the performance of machine learning algorithms on temporally dynamic data. Using analytical derivations and Monte Carlo analyses, I contribute to the literature on time series analysis by helping avoid spurious regressions and improving accessibility to time series analysis. I provide a Stata program in which I calculate sample-specific critical values for the Kwiatkowski-Phillips-Schmidt-Shin (KPSS) test, I recommend the use of a general modeling strategy that avoids the need for unit root testing, and I demonstrate how the ordinary least squares (OLS) estimator outperforms more complex machine learning algorithms when using temporally dynamic data.