William Hakes- IEEE Article Summary II
1) “A Comparison of Linear and Nonlinear Statistical Techniques in Performance Attribution” IEEE Transactions on Neural Networks, Ngai Hang Chan and Christopher R. Genovese. Vol. 12, NO. 4, July 2001.
2) Ngai Hang Chan and Christopher R. Genovese. Dept of Statistics, Chinese University of Hong Kong Shatin, and Dept of Statistics, Carnegie Mellon University, respectively.
3) Are nonlinear models more successful at predicting performance vs. traditional linear models?
4) Applied Statistical: This article takes some real world data, and then applies it using various sophisticated quantitative techniques.
5) Linear models have been around for some time. In the area of performance measurement, predicting performance using various factors, these linear models are relatively easy to understand and computationally efficient. The question arises, though- are these reasons sufficient to justify their use? Or should we use better methods, more suited to the data, particularly since we now have better computing power? This paper demonstrates that nonlinear models are a step in the right direction. The results were surprising- the most sophisticated model (Neural networks) was the least successful. Implications are discussed, including the need for more factors as inputs as well as a more sophisticated Neural Network equipped with a learning algorithm.
6) The article might appeal to both academics and practitioners interested in the study of finance as well as general quantitative methods. Most likely, though, academics will find this more suitable, as it is just an introduction of many subsequent tests to come.
7) I would like to publish an article such as this, comparing the techniques across a much larger and less biased time horizon.