Identification of Spikes in Time Series

TL;DR

This study evaluates various spike detection methods in time series data through simulations and real-world violence rate analysis, finding Kalman filtering and smoothing to be the most effective.

Contribution

It systematically compares multiple spike detection techniques in a simulation setting and applies the best method to real-world data.

Findings

Kalman filtering and smoothing outperformed other methods in sensitivity and specificity.

The best method successfully identified spikes in violence rates across California cities.

Simulation results guide practical spike detection in social science time series.

Abstract

Identification of unexpectedly high values in a time series is useful for epidemiologists, economists, and other social scientists interested in the effect of an exposure spike on an outcome variable. However, the best method to identify spikes in time series is not known. This paper aims to fill this gap by testing the performance of several spike detection methods in a simulation setting. We created simulations parameterized by monthly violence rates in nine California cities that represented different series features, and randomly inserted spikes into the series. We then compared the ability to detect spikes of the following methods: ARIMA modeling, Kalman filtering and smoothing, wavelet modeling with soft thresholding, and an iterative outlier detection method. We varied the magnitude of spikes from 10-50% of the mean rate over the study period and varied the number of spikes inserted from 1 to 10. We assessed performance of each method using sensitivity and specificity. The Kalman filtering and smoothing procedure had the best overall performance. We applied Kalman filtering and smoothing to the monthly violence rates in nine California cities and identified spikes in the rate over the 2005-2012 period.