How It Works
Financial markets are highly complex systems, with thousands of prices moving in response to changes in incoming information on dozens of variables such as economic growth, inflation expectations, energy prices, corporate borrowing costs, earnings estimates and emerging market volatility. Traditional linear models of asset prices, such as multiple regression, have been unable to deliver valid results because the predictor variables are highly correlated (multicollinearity) and sometimes non-stationary.
QI’s processes pick out data relevant to a security, use algorithms to derive the combinations of variables that best explain price movements (“components”) and then calculate the price that is consistent with recent relationships. Independent associations are stripped out and the sensitivity of securities to macro factors is calculated.These processes occur in real time throughout the day and cover thousands of instruments.
The QI process has been designed to overcome the problems of overfitting, multicollinearity and non-stationarity. The process has been developed in conjunction with Professor Michael Hobson, Vice-Master, Trinity Hall, University of Cambridge. Professor Hobson is a leader in the fields of Bayesian inference, machine learning and Principal Component Analysis.