Abstract: The extensive coverage of suicides in the media has long been considered a trigger for copycat suicides. However, evidence of such an effect is indirect and, thus, inconclusive. Here, we propose a flexible threshold autoregressive model to examine whether suicides reported in the media influence actual suicides and, thus, identify a possible copycat effect. In particular, we employ a penalized smoothing least squares estimator to conveniently estimate the parameters and unknown functions of the proposed model. We evaluate the performance of the proposed method using simulation studies, and examine the asymptotic behavior of the corresponding estimators under mild regularity conditions. Lastly, we apply our model to investigate the relationship between the daily suicide incidence and the number of suicides reported in a top-selling tabloid newspaper in Hong Kong between January 2002 and December 2006. Our results identify a copycat suicide effect due to excessive media reporting, as well as the threshold number of reports that may trigger such an effect.

Key words and phrases: Autoregressive model, copycat suicide effect, penalized smoothing least squares estimator, threshold model, time series data.