The Impact of COVID-19 Crisis on Stock Markets’ Statistical Complexity

Figure 3

Bayesian network for a compound measure of complexity and its arcs according to the strength of dependencies (full sample). Notes: Tabu search score-based structure-learning algorithm. This is a modified hill-climbing algorithm able to escape local optima by selecting a network that minimally decreases the score function. The procedure of Friedman et al. [39] measures the degree of confidence in a particular graphical feature of a Bayesian network by generating multiple network structures by applying nonparametric bootstrap to the data and estimating the relative frequency of the feature of interest. The parameterisation follows Imoto et al. [40]. One thousand bootstrap replicates are performed. The figure displays a measure of confidence or strength for each arc while keeping the rest of the network structure fixed. The minimum strength required for an arc to be included in the averaged network is set to 0.50.