A1216
Title: Crypto analysis with extreme value theory
Authors: M Cristina Miranda - University of Aveiro (Portugal) [presenting]
Manuela Souto de Miranda - University of Aveiro (Portugal)
Conceicao Amado - Universidade de Lisboa (Portugal)
Ivette Gomes - FCiencias.ID, Universidade de Lisboa and CEAUL (Portugal)
Abstract: Financial time series data often exhibit clusters of high or low values. The extreme, rare-event behavior inherent to this data type is typically modeled using extreme value distributions. When dealing with dependent data, an additional parameter, the extremal index, must be considered. This parameter can be interpreted as the inverse of the limit mean size of clusters of high values and also as the proportion of non-zero inter-exceedance times. Estimating the extremal index is crucial for generating more accurate predictions. The purpose is to compare the proportion extremal index estimator with other widely recognized estimators from the literature, evaluating their performance when applied to real samples of cryptocurrency data and providing confidence intervals for the parameter.
