Leverage in Regression Models with MA(1) Errors

오차항이 MA(1) 과정을 따르는 회귀모형에서의 Leverage

This paper investigates the effect of individual observations in regression models with MA(1) errors through the 'hat matrix' It shows that the first observation has the largest hat matrix diagonal component for $\theta$＜0 in the regression model with an intercept. This provides additional evidence for retaining the first observation in performing estimation in this setting. When the regression model goes to the origin and the independent variable has a deterministic trend, the last observation has the greatest leverage for │$\theta$│＜1 and may have potentially large impact on parameter estimation.