No description available.

No description available.

No description available.

No description available.

No description available.

We propose a new framework for the analysis of hedge funds and the modelling of their per- formance. This approach is based on our finding that the investment styles declared by fund managers are unreliable and are uninformative about fund performance. The framework has been designed to cluster hedge funds using an innovative technique based on artificial neural networks. The framework is tested on a large selection of different data sets and the resulting clusters provide the best matches with known solutions when compared to existing methods. The framework applied to hedge funds is a valuable fund management tool as funds are categorised in terms of their performance alone. The universe of hedge fund is best represented by 12 clusters, each corresponding to a different pattern of returns. The clusters are shown to be stable over time and this indicates that the framework will continue to be a reliable representation of the hedge fund market after the Global Credit Crisis. We demonstrate that the framework allows the construction of a more efficient frontier of portfolios by diversifying across the clusters identified rather than using investment styles. A multi-factor model is built using these clusters; it dominates all comparable models in the ability to explain fund returns. It allows managerial skill to be.estimated more accurately, leading to a better valuation of hedge funds and fund of hedge funds.

This thesis explores pricing models for interest rate markets. The model used to ':describe the short rate is based on the discontinuous shot noise process. As a consequence the market is incomplete, meaning that not all securities contingent on the short rate can be replicated perfectly with a dynamically adjusted portfolio of a bond and cash. This framework is still consistent with the absence of arbitrage as evidenced by the existence of an equivalent martingale measure. This measure is not unique, however, due to the incompleteness of the market. Two approaches to pricing contingent claims are pursued. The first, risk-neutral pricing, evaluates the expected value of the pay-off at expiration under an equivalent martingale measure. A parameterized class of martingales, based on the Esscher transform, allows for the definition of a flexible set of equivalent martingale measures and results in a formula for the conditional joint Laplace transform of the short rate and its time-integral. The pricing formula for a discount bond follows trivially from these results. A method for pricing a European call option is also proposed, requiring numerical inversion of the aforementioned Laplace transform. The second approach, mean-variance hedging, addresses the incompleteness of the market. A contingent claim is priced by forming a portfolio of a bond and cash. The portfolio is dynamically updated to .mimic the pay-off of the claim at expiration. The replicating portfolio is restricted to be self-financing and predictable. This approach leads to a closed-form pricing formula for a discount bond and formulae for European call and put options, requiring the numerical Laplace inversion methods mentioned above. All this is in the context of a discrete-time model that includes as a special case a discrete-time version of the shot noise process.

No description available.

No description available.

No description available.