This paper discusses the relationship between the least square estimate (LSE) and the best linear unbiased estimate (BLUE) of the mean vector in a linear model with a singular covariance matrix. The identical sufficient and necessary conditions of these t