| Mostafa Zahed, Maryam Skafyan |
This study applies multivariate time series analysis to compare age-standardized death rates for lung and colorectal
cancer in the United States from 1950 to 2021, using data from Our World in Data. Lung cancer, the leading cause of
cancer mortality, and colorectal cancer, the second most common, are analyzed to uncover long-term trends, potential
seasonal patterns, and interdependencies. Vector Autoregression (VAR) models capture joint dynamics, with
cointegration analysis to detect long-run equilibrium relationships and Granger causality tests to assess predictive
relationships between the series. Impulse response functions and variance decomposition explore how shocks in one
series affect the other. The study also generates forecasts for future years, offering insights into potential trends in cancer
mortality. Hypothesis testing assumes stationarity (verified via Augmented Dickey-Fuller tests), no residual
autocorrelation (checked with Ljung-Box tests), and, where applicable, normality of residuals (assessed using
Jarque-Bera tests). Structural breaks, potentially linked to policy changes, are examined using Chow or Bai-Perron tests.
Findings will inform public health strategies, highlighting advanced time series methods for addressing cancer mortality
challenges. |