Quantitative Porfolio Construction - Practical Considerations
Overview
Robust
Quantitative Considerations
Many quantitative models are predicated on a set of assumptions, sometimes assuptions that are quite tenuous.
(As an example, think asset returns being
normally distributed)
When relying on
a model, it is necessary to understand how violations of assumptions can affect the results, and my require some monitoring
to make sure that model violations are small, including a plan for dealing when violations become material.
Many strategies make distributional assumptions about its variables. That is, they assume that the variable follows a certain
distribution,
and only the parameters of the distribution need to be measured. In contrast, a robust strategy might
assume that the distribution itself is random, and any parameters used in the model are analyst best guesses.
(see
bayesian statistics for instance)
Many strategies are sound theoretically, but they depend on being able to measure certain parameters. Often these paremeters
are impossible to measure with any degree of certainty.
(see
returns vs volatility for instance)
The best strategies are designed to be robust respect to its parameters.
This means either using only parameters that can be measured reliably, or to build the strategy with low sensitivity to the
unreliable parameters.
Robust Strategies
