Q1. Let Y = Xβ + \varepsilon, where Y is n imes 1 vector of response, X is n imes k (n > k) matrix of predictors, β is k imes 1 vector of unknown coefficients to be estimated and \varepsilon is n imes 1 vector of random errors. Further \varepsilon is N(0, σ^2 I). The distribution of \hatY is :
- (a) N(Xβ, σ^2)
- (b) N(Xβ, X(X'X)-1 X'σ^2)
- (c) N(Xβ, (X'X)-1 σ^2)
- (d) N(Xβ, (X'X)-1 X'σ^2)