Multivariate Models

Multivariate Distributions in Non-Stationary Complex Systems I: Random Matrix Model and Formulae for Data Analysis ArXiv ID: 2412.11601 “View on arXiv” Authors: Unknown Abstract Risk assessment for rare events is essential for understanding systemic stability in complex systems. As rare events are typically highly correlated, it is important to study heavy-tailed multivariate distributions of the relevant variables, especially in the presence of non-stationarity. We use a generalized scalar product between correlation matrices to clearly demonstrate this non-stationarity. Further, we present a model that we recently put forward, which captures how the non-stationary fluctuations of correlations make the tails of multivariate distributions heavier. Here, we provide the resulting formulae including Gaussian or Algebraic features. Compared to our previous results, we manage to remove in the Algebraic cases one out of the two, respectively three, fit parameters which considerably facilitates applications. We demonstrate the usefulness of these results by deriving joint distributions for linear combinations of amplitudes and validating them with financial data. Furthermore, we explicitly work out the moments of our model distributions. In a forthcoming paper we apply the model to financial markets. ...

Vector Autoregression in Cryptocurrency Markets: Unraveling Complex Causal Networks ArXiv ID: 2308.15769 “View on arXiv” Authors: Unknown Abstract Methodologies to infer financial networks from the price series of speculative assets vary, however, they generally involve bivariate or multivariate predictive modelling to reveal causal and correlational structures within the time series data. The required model complexity intimately relates to the underlying market efficiency, where one expects a highly developed and efficient market to display very few simple relationships in price data. This has spurred research into the applications of complex nonlinear models for developed markets. However, it remains unclear if simple models can provide meaningful and insightful descriptions of the dependency and interconnectedness of the rapidly developed cryptocurrency market. Here we show that multivariate linear models can create informative cryptocurrency networks that reflect economic intuition, and demonstrate the importance of high-influence nodes. The resulting network confirms that node degree, a measure of influence, is significantly correlated to the market capitalisation of each coin ($ρ=0.193$). However, there remains a proportion of nodes whose influence extends beyond what their market capitalisation would imply. We demonstrate that simple linear model structure reveals an inherent complexity associated with the interconnected nature of the data, supporting the use of multivariate modelling to prevent surrogate effects and achieve accurate causal representation. In a reductive experiment we show that most of the network structure is contained within a small portion of the network, consistent with the Pareto principle, whereby a fraction of the inputs generates a large proportion of the effects. Our results demonstrate that simple multivariate models provide nontrivial information about cryptocurrency market dynamics, and that these dynamics largely depend upon a few key high-influence coins. ...