A0449
Title: Test of serial dependence or cross dependence for time series with underreporting
Authors: Keyao Wei - National University of Singapore (Singapore) [presenting]
Yingcun Xia - National University of Singapore (Singapore)
Abstract: Testing the serial dependence of a time series or cross dependence of two-time series is essential in time series analysis. Many efficient methods have been proposed for the test when the data is accurately recorded. However, the observed data often systematically deviate from the actual values, a common example being data underreporting in social sciences, ecology, and epidemiology. For these data, it may not be possible to directly make correct inferences using traditional statistical tests. New tests are introduced by using the lag differences of the observed time series, and the statistical consistency of the tests is proven. Further, a block bootstrap method is used to mimic the asymptotic distribution of the test statistics. Numerical experiments show that the proposed tests perform better than existing methods in the case of underreporting. This method has successfully detected important factors for dengue transmission and cardiovascular diseases.
