Master Degree Project Presentation: An Empirical Study on SPY

Abstract: This project applies logistic regression to predict the daily direction of the S&P 500 using lagged returns, lagged directional indicators, and technical indicators such as RSI and MACD. Three models are estimated: a Full Model, a Lag-Only Model, and a Technical-Only Model, these models are then compared using the Akaike Information Criterion (AIC). The Full Model achieved the lowest AIC value, indicating it had the best overall fit. Three residual diagnostics were run to assess the model, these include: standardized deviance residuals, leverage measures, and Anscombe residuals.

The in-sample confusion matrix acquired an accuracy of 73%, suggesting that the model contains some predictive power. Combining that with a white noise simulation that achieved 50% accuracy gives reason that logistic regression can identify weak but meaningful patterns in daily market movements.

The presentation is taking place on Zoom (meeting ID: 699 250 9213)