Data transformations are commonly used across statistics to transform data distributions into distributions with properties that make them more user friendly. In time-series, stationarity is one of the most common assumptions that is violated because the mean and variance are time dependent....

We examine machine learning and factor-based portfolio optimization. We find that factors based on autoencoder neural networks exhibit a weaker relationship with commonly used characteristic-sorted portfolios than popular dimensionality reduction techniques. Machine learning methods also lead to...

I show that an important no-arbitrage consistent but costly collateral rental yield contributes to about two-thirds of the standard CIP violations. I measure this yield using two approaches applied to short- and long-term CIP horizons. First, I assume that the yield is observable and proxy it...

We look at the financial markets as represented by a network of agents similar to bond percolation models in physics or epidemiology models. We aim to figure out how an agent based network model can cause perturbations that can cause failures of the traditional economic theory, specifically the...

In the paper we introduce an empirical approximation of the log-optimal investment strategy that guarantees an almost optimal growth rate of investments. The proposed strategy also considers the effects of portfolio rearrangement costs on growth optimality and advises a suboptimal portfolio for...

Using the Tsay (1988) outlier identification methodology on daily log-returns of 16 commodity spot price series and 25 commodity index series, this study assesses the impact significant and unexpected news announcements had on volatility between January 1, 1997 and December 31, 2007. Results...

• Various statistics of future price level excursions are driving decision making of both sophisticated and amateur traders, although the latter are not necessarily aware of it. Such statistics are of paramount importance in intelligent placement of limit orders and monitoring their...

Demonstration that noise filtered correlation matrices can be used for early detection of a regime change in temporal behavior of securities. This demonstration was carried out for a portfolio of 40 S&P500 securities with just two, randomly chosen, securities undergoing a deliberately arranged...

Demonstration that in-sample Markowitz type mean-variance optimization, carried out with noise filtered covariance matrices, results in asset allocation that leads to 2-3 times increase of the Sharpe ratio compared to the same optimization carried out without noise filtering.Demonstration of 2-3...

Demonstration of the extraordinary out-of-sample stability of noise-filtered correlation matrices measured in terms of simple Euclidean distance. This measure decreases 4-5 times post noise filtering implying 4-5 times improved stability.Study of the out-of-sample stability of noise-filtered...