Dealing With Heteroskedasticity Within the Modeling of the Quality of Life of Older People

Abstract

Using the estimation method of ordinary least squares leads to unreliable results in the case of heteroskedastic linear regression model. Other estimation methods are described, including weighted least squares, division of the sample and heteroskedasticity-consistent covariance matrix estimators, all of which can give estimators with better properties than ordinary least squares. The methods are presented giving the example of modelling quality of life of older people, based on a data set from the first wave of the COURAGE-Poland study. The comparison of estimators and their practical application may teach how to choose methodologically the most appropriate estimation tool after detection of heteroscedasticity.

Estimating Weibull Parameters Using Least Squares and Multilayer Perceptron vs. Bayes Estimation

“…Manual approaches include the ordinary least squares [15,16], unbiased good linear estimators [17], and weighted least squares [18]. Computational methods include maximum likelihood estimation [19], the moments estimation method [20], Bayesian approach [21], and least-squares estimation with particle swarm optimization [22].…”