Title: Long memory conditional heteroscedasticity in count data
Authors: Manuel Stapper - WWU Muenster (Germany) [presenting]
Abstract: A new class of long memory integer-valued processes is introduced, which are adaptations of the well-known FIGARCH and HYGARCH processes to a count data setting. Statistical properties of the models are provided and it is shown via simulation that reasonable parameter estimates are easily obtained via conditional maximum likelihood estimation. An asymptotic test is derived and used to test for restrictions. To illustrate the practical importance of the models, an empirical application with financial transaction data is performed. For this purpose, high frequency data is collected and the number of price changes in 60-second intervals used as time series.