Significant conclusions on the dynamics of the market graph

This work addresses the problem of detecting the dynamics of the market graph. The key question is whether observed changes in the market graph are systematic or can be explained by random fluctuations in correlation estimates. A method is proposed to obtain statistically significant conclusions about the presence of market graph dynamics. This method builds on a recently developed approach for identifying sets of statistically significant inferences regarding the existence or absence of correlations of a specific strength between each pair of stocks. According to the method a decision on the presence of  dynamics is made if at least one pair of stocks in one observation period is classified as having a significant correlation while in another period it is classified as lacking a significant correlation. The proposed approach allows identifying pairs of stocks responsible for the market graphs dynamics, pair whose correlation changes do not lead to such dynamics and an area of uncertainty. Examples of applying the method to analyze the dynamics of the market graph constructed for key stocks of the Russian stock market are provided.  
This is a joint work with Kodanov P.A., Koldanov A.P., Tsygankova E.P.