STAT2004J – Linear Modelling Tutorial 6
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
STAT2004J – Linear Modelling
Tutorial 6
Question 1.
Consider the linear regression model in which responses Yi are uncorrelated and have expecta- tions β0 + β1Xi and common variance σ2 (i = 1, ..., n). First, show that
Hence, show that Cov = 0, and that the residuals deined by
satisfy
where
Which of the residuals has the largest variance?
Show that the average of the variance of the ˆ(ε)i’s is close to σ2 for large n.
Question 2.
Suppose that, given the values Xi of an explanatory variable X, the responses Yi are uncorre-lated and have expectations β0 + β1Xi (i = 1, ..., n). In this exercise, we allow the variances of Yi ’s to be diferent for diferent i’s, i.e., we assume the model is heteroscedastic. For every i, we denote Var. In this case, for every i, the normalized errors will all have unit variance and zero mean. So a suitable method for estimating the parameters β0 and β1 under these assumptions is the weighted least squares, which minimizes the weighted sum of squares
with respect to β0 and β1. In other words
Show that the weighted least squares estimates and satisfy the two equations
2024-01-08