Should We Demean the Data?
Annals of Economics and Finance, 2015, vol. 16 (1), 163-171
Posted: 27 Feb 2016
Date Written: February 25, 2015
Abstract
The sample average is an unbiased estimator of the population mean, so it may seem innocuous that for estimating model parameters that do not involve the population mean, the data can be demeaned first. Using a first-order moving average (MA) model for example, we derive the analytical approximate biases of the quasi maximum likelihood estimators (QMLEs) based on the original and demeaned data. The bias results indicate that the QMLEs can behave quite differently in finite samples and it is not always advisable to demean the data if the MA parameter is of primary interest to estimate.
Keywords: Demean, Moving Average, Bias
JEL Classification: C22, C51
Suggested Citation: Suggested Citation