时间序列分析的Gauss-Markov

If Unbiased :

1. Linear
   y_t = α + β₁x₁ + β₂x_2t+u_t

2. Zero Conditional Mean of Error
   E[u_t|x_jk]=0

3. No Perfect Collinearity

4. Homoscedasticity
   var(u_t|x_jk)=δ²

5. No Serial Correlation
   cov(u_t,u_s|x_jk)=0