Detection of Patches of Outliers in Stochastic Volatility Processes.

Autores

  • Anderson C.O. Motta
  • Luiz K. Hotta Department of Statistics, University of Campinas

DOI:

https://doi.org/10.11606/issn.2316-9028.v8i2p169-191

Palavras-chave:

Patches of outliers, Stochastic volatility, Detection of outliers, Markov chain Monte Carlo.

Resumo

Because the volatility of nancial asset returns tends to arrive in clusters, it is quite likely that outliers appear in patches. In this case, most of the statistical tests developed to detect outliers have low power. We propose to use the posterior distribution of the size of the outlier and of the probability of the presence of an outlier at each observation to detect and estimate the outlier. This sampling algorithm is an adapted version of the algorithm proposed by Justel et al. (2001) for autoregressive time-series models. Our proposed sampling procedure is applied to a simulated sample according to the stochastic volatility, a sample of the New York Stock Exchange daily returns, and a sample of the Brazilian S~ao Paulo Stock Exchange daily returns.

Downloads

Não há dados estatísticos.

Biografia do Autor

Luiz K. Hotta, Department of Statistics, University of Campinas

Department of Statistics, University of Campinas

Downloads

Publicado

2014-12-12

Como Citar

Motta, A. C., & Hotta, L. K. (2014). Detection of Patches of Outliers in Stochastic Volatility Processes. São Paulo Journal of Mathematical Sciences, 8(2), 169-191. https://doi.org/10.11606/issn.2316-9028.v8i2p169-191

Edição

Seção

Artigos