RPESE: Risk and Performance Estimators Standard Errors with Serially Dependent Data

Authors

Anthony-Alexander Christidis, University of British Columbia
R Douglas Martin, University of Washington

Date of this Version

12-2021

Document Type

Article

Citation

The R Journal (December 2021) 13(2); Editor: Dianne Cook

Comments

Copyright 2021, The R Foundation. Open access material. License: CC BY 4.0 International

Abstract

The R package RPESE (Risk and Performance Estimators Standard Errors) implements a new method for computing accurate standard errors of risk and performance estimators when returns are serially dependent. The new method makes use of the representation of a risk or performance estimator as a summation of a time series of influence-function (IF) transformed returns, and computes estimator standard errors using a sophisticated method of estimating the spectral density at frequency zero of the time series of IF-transformed returns. Two additional packages used by RPESE are introduced, namely RPEIF which computes and provides graphical displays of the IF of risk and performance estimators, and RPEGLMEN which implements a regularized Gamma generalized linear model polynomial fit to the periodogram of the time series of the IF-transformed returns. A Monte Carlo study shows that the new method provides more accurate estimates of standard errors for risk and performance estimators compared to well-known alternative methods in the presence of serial correlation.