CMStatistics 2021: Start Registration
View Submission - CFE
A0284
Title: Hypotheses testing in mixed-frequency volatility models: A bootstrap approach Authors:  Vincenzo Candila - University of Salerno (Italy) [presenting]
Lea Petrella - Sapienza University of Rome (Italy)
Alberto Arcagni - Sapienza University of Roma (Italy)
Abstract: It is widely recognized that standard likelihood-based inference suffers from the presence of nuisance parameters. This problem is particularly relevant in the context of Mixing-Data Sampling (MIDAS) methods applied to volatility modeling. In this framework, the volatility can be decomposed into two components, one varying daily and another varying according to the (lower) frequency of the additional volatility determinant. The MIDAS methods estimate the weights associated with each lagged realization of the low-frequency variable. A problem arises when the interest is on testing the whole impact of the low-frequency variable because, under the null hypothesis of no influence, the weight parameters are not identifiable. From this aspect, the weight parameters can be seen as nuisance parameters. This situation interferes with the asymptotic distribution of the common statistical tests employed to evaluate the significance of all the model's parameters. In order to overcome this problem, a bootstrap likelihood ratio (BLR) test is proposed, simulating the likelihood ratio test distribution. Using a Monte Carlo experiment, the proposed BLR test presents considerably good performances in terms of the test's size and power, generally better than the standard likelihood ratio test.