CMStatistics 2016: Start Registration
View Submission - CMStatistics
B0754
Title: Limit theorems for residuals from VARMAX models with potentially serially correlated errors Authors:  Benjamin Holcblat - University of Luxembourg (Luxembourg) [presenting]
Steffen Gronneberg - BI Norwegian Business School (Norway)
Abstract: In time series analysis, numerous inference procedures need to use residuals instead of the unobserved error terms. Allowing for serial correlation between error terms, we prove limit theorems for partial sums of smooth functions of VARMAX residuals. Process limits based on residuals do not correspond to process limits based on the error terms, generally. However, the former correspond to the process limits of average-corrected error terms. Applications of our results include estimation of log-GARCH models, generalization of Jarque-Bera normality tests, and CUSUM-type of tests.