给定一组样本 ( X i , Y i ) (X_i,Y_i) (Xi,Yi),模型 Y i = β 0 X i + μ i Y_i=\beta_0X_i+\mu_i Yi=β0Xi+μi的离差形式为:
y i = Y i − Y ‾ = β 1 x i + ( μ i − μ ‾ ) y_i=Y_i-\overline Y=\beta_1x_i+(\mu_i-\overline\mu) yi=Yi−Y=β1xi+(μi−μ)
样本回归函数的离差形式:
y ^ i = β ^ 1 x i \hat y_i=\hat \beta_1x_i y^i=β^1xi
所以
∑ e i 2 = ∑ ( y i − y ^ i ) 2 = ∑ [ ( β 1 − β ^ 1 ) x i + ( μ i − μ ‾ i ) ] 2 = ∑ ( β 1 − β ^ 1 ) 2 x i 2 + ∑ ( μ i − μ ‾ ) 2 + 2 ∑ ( β 1 − β ^ 1 ) x i ( μ i − μ ‾ i ) ] = ∑ ( β 1 − β ^ 1 ) 2 x i 2 + ∑ ( μ i − μ ‾ ) 2 − 2 ∑ ( ∑ k i μ i ) x i ( u i − μ ‾ ) = ∑ ( β 1 − β ^ 1 ) 2 x i 2 + ∑ ( μ i − μ ‾ ) 2 − ∑ x i μ i ∑ k i μ i + 2 μ ‾ ∑ x i ∑ k i μ i = ∑ ( β 1 − β ^ 1 ) 2 x i 2 + ∑ ( μ i − μ ‾ ) 2 − 2 ∑ x i μ i ∑ x i μ i ∑ x i 2 \sum e_i^2=\sum(y_i-\hat y_i)^2=\sum[(\beta_1-\hat \beta_1)x_i+(\mu_i-\overline \mu_i)]^2\\\quad\\=\sum(\beta_1-\hat\beta_1)^2x_i^2+\sum(\mu_i-\overline \mu)^2+2\sum(\beta_1-\hat\beta_1)x_i(\mu_i-\overline \mu_i)]\\\quad\\=\sum(\beta_1-\hat\beta_1)^2x_i^2+\sum(\mu_i-\overline \mu)^2-2\sum (\sum k_i\mu_i)x_i(u_i-\overline \mu)\\\quad\\= \sum(\beta_1-\hat\beta_1)^2x_i^2+\sum(\mu_i-\overline \mu)^2-\sum x_i\mu_i\sum k_i\mu_i+2\overline\mu\sum x_i\sum k_i\mu_i\\\quad\\= \sum(\beta_1-\hat\beta_1)^2x_i^2+\sum(\mu_i-\overline \mu)^2-2\sum x_i\mu_i\frac{\sum x_i\mu_i}{\sum x_i^2} ∑
笔记:干扰项方差的无偏估计
最新推荐文章于 2024-04-01 21:44:57 发布