Extended Skew Kalman Filters for COVID-19 Pandemic State Estimation

Abstract

This research studies the long-term behavior monitoring of the COVID-19 pandemic through estimation with nonzero skewness. The COVID-19 data may contain outliers that could result in inaccurate estimation using traditional Gaussian Kalman filtering methods due to asymmetry in the posterior distribution. This paper aims to address this issue by employing a skewed Kalman filter (SKF) that considers skewness in the relevant quantities. A novel epidemiological model is introduced, and the efficient Bayesian inference algorithm nested sampling is utilized to determine the posterior distribution of the time-varying epidemiological model parameters. The study aims to estimate the number of active cases and deaths in Saudi Arabia over long-term as well providing estimates of the hidden states. Finally, the results are compared with the deterministic pandemic model and the skewed Kalman filter estimates.