Suppose that we have the following linear model $\mathbb{E}[Y_{i}]=b_{0}+b_{1}(x_{i}-\bar{x})$ where $i=1,...,n$ and $\bar{x}$ is the mean of $x_{i}$ and $Y_{i}$ are uncorrelated with constant variance $\sigma^{2}$. We derived that the least squares estimators of $b_{0}$ and $b_{1}$ are the following

$\hat{b_{0}} = \bar{Y}$ and $\hat{b_{1}}=\frac{\sum_{i}(x_{i}-\bar{x})Y_{i}}{\sum_{i}(x_{i}-\bar{x})^{2}}$

We also know from the covariance matrix of the estimators that $\mathbb{E}[\hat{b_{0}}\hat{b_{1}}]=0$ are uncorrelated.

Now we are given a different linear function $\tilde{b}=\sum_{i}b_{i}Y_{i}$ where $b_{i}$ are constants. And we are also given the mean square error of $\tilde{b}$ which is

$MSE(\tilde{b}) = var(\tilde{b})+(\mathbb{E}(\tilde{b})-b_{0})^{2}$.

My goal is to prove that $MSE(\tilde{b})$ is independent of $b_{1}$ if and only if $\sum_{i}b_{i}(x_{i}-\bar{x})=0$

Thus we have to prove that

i) $\mathbb{E}[MSE(\tilde{b})b_{1}]=\mathbb{E}[MSE(\tilde{b})]\mathbb{E}[b_{1}]\Rightarrow \sum_{i}b_{i}(x_{i}-\bar{x})=0$

ii) with $\sum_{i}b_{i}(x_{i}-\bar{x})=0 \Rightarrow \mathbb{E}[MSE(\tilde{b})b_{1}]=\mathbb{E}[MSE(\tilde{b})]\mathbb{E}[b_{1}]$

But I am struggling on how to prove it.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.