$X_1,\ldots,X_n$ are non-random constants. We observe random variables $Y_i,\ldots,Y_n$

that are independent but not identically distributed

$Y_i\sim N(\alpha+\beta X_i,\sigma^2)$ where $\alpha,\beta$ and $\sigma^2$ are unknown parameters

Question: Find the asymptotic joint distribution of the MLE.

The asymptotic distribution of the MLE estimator should be:

$$\sqrt n (\hat{\theta}-\theta)\sim N(0,\frac{1}{I_{\theta}})$$

and the MLE of $\alpha,\beta$ and $\sigma^2$ should be:

$$\hat{\alpha}=(\bar{y}-\hat{\beta} \bar{X})$$


$$\widehat{\sigma^2}=\frac{\sum_i^n(y_i-\hat{\alpha}-\hat{\beta} x_i)^2)}{n}$$

But how do I calculate the asymptotic joint distribution of the MLE?


1 Answer 1


There is a multivariate version of the convergence in distribution that you mentioned.

With parameter $\theta = (\alpha, \beta, \sigma^2)$ and MLE $\hat{\theta} = (\hat{\alpha}, \hat{\beta}, \hat{\sigma^2})$, under certain conditions we have $$\sqrt{n}(\hat{\theta} - \theta) \overset{d}{\to} N(0, I^{-1})$$ where $I$ is the Fisher information matrix.

  • $\begingroup$ This omits to mention that $\widehat\alpha,\widehat\beta$ are correlated, and does not explain why $\widehat{\sigma^2}$ is independent of $(\widehat\alpha,\widehat\beta). \qquad$ $\endgroup$ Commented Nov 7, 2020 at 1:50

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .