The Dependent Error Regression Smoothing Approach For Semi-parametric Evoked Potential Estimation
Michael G. Schimek · 2005
In this paper we discuss statistical concepts for the estimation of evoked potential signals. Classical parametric techniques are inappropriate as they do not allow the researcher to separate the signal from the noise. Recent non-parametric approaches have overcome this major drawback but require the unrealistic assumption of white noise errors. We propose semiparametric signal estimation by the Dependent Error Regression Smoothing (DERS) approach, which relaxes the white noise assumption in that also prespecified autoregressive and moving average error structures can be considered. The application of DERS is demonstrated on a visual evoked potential series.