Small Sample Inferences on the Sharpe Ratio

This work deals with statistical inferences on the "Sharpe Ratio" (SR) based on small samples. We have considered point estimation, interval estimation, as well as hypothesis testing, assuming that a random sample is available from a normal distribution. Further, we study the robustness of our inferential methods when the data is thought to have come from other nonnormal distributions but is mistakenly modeled by the normal distribution. Results from a comprehensive simulation study have been provided to justify our observations and recommendations. Among other things, we have proposed a new estimator of SR that performs much better than the commonly used maximum likelihood estimator. Finally, some mutual fund datasets have been used for demonstration purposes to estimate SR in order to assess their monthly return performances.