We formulate a model for the off-line estimation of a changepoint in a network setting. The framework naturally allows the parameter space (network size) to grow with the number of observations. We compute the signal-to-noise ratio detectability threshold, and establish the dependence of the rate of convergence and asymptotic distribution on the network size and parameters. In addition, we show that inference can be adaptive, i.e. asymptotically correct confidence intervals can be computed based on the data. We apply the method to the question of whether US Congress has abruptly become more polarized at some point in recent history.
|Original language||English (US)|
|Title of host publication||Stochastic Models, Statistics and Their Applications, 2015|
|Editors||Ewaryst Rafajłowicz, Krzysztof Szajowski, Ansgar Steland|
|Publisher||Springer New York LLC|
|Number of pages||9|
|State||Published - 2015|
|Event||12th Workshop on Stochastic Models, Statistics and their Applications, 2015 - Wrocław, Poland|
Duration: Feb 16 2015 → Feb 20 2015
|Name||Springer Proceedings in Mathematics and Statistics|
|Other||12th Workshop on Stochastic Models, Statistics and their Applications, 2015|
|Period||2/16/15 → 2/20/15|
Bibliographical noteFunding Information:
E.Y.’s research was partially supported by US NSF grant DMS-1204311. M.B.’s research was partially supported by US NSF DMS-1007751, US NSA H98230-11-1-0166, and a Sokol Faculty Award, University of Michigan. G.M.’s research was partially supported by US NSF DMS-1228164 and US NSA H98230-13-1-0241. The authors thank the referees for helpful comments.
© Springer International Publishing Switzerland 2015.