![IconGems-Print](images/IconGems-Print.png)
65 – Forecasting and Time Series Estimation
Efficient Non-Parametric Spectral Density Estimation with Randomly Censored Data
Jiaju Wu
The University of Texas at Dallas
Spectral Density estimation is a well known problem for a directly observed time series. The literature on spectral density estimation for a right-censored time series is next to none. A data-driven spectral density estimator for a right censored time series is suggested. This estimator adapts to unknown smoothness of the spectral density and unknown distribution of a censored random variable. Asymptotic upper bound of the mean integrated squared error (MISE) of the proposed estimator is obtained. The estimator is studied via simulated examples.