Bootstrap based measurement of serial correlation in time series objects
MetadataShow full item record
Serial correlation is a fundamental problem in time series analysis. Since financial data are collected every few minute and due to the large-scale of data, finding the correlation of time series data accurately becomes a crucial problem. This thesis explores the monthly correlation of time series data when daily values are observed. In this case, we call daily values within a month a time series object, and the goal of this study is to measure the correlation in time series objects. Traditionally, researchers often take the last day value, but this approach may lose information within the month. This thesis proposes an approach based on resampling techniques, bootstrap and wild bootstrap, and compare their finite sample performances with the that of the last day approach.