1

I have a no-intercept relationship:

$$y_{i} = \beta_{1}x_{i} + \varepsilon_{i}$$

where $\varepsilon_{i} \sim \text{ iid } \mathcal{N}(0, \sigma^{2})$, and $i = 1, \dots, n$.

How do I derive $\hat{\beta}_{1}$, the least-squares estimator of $\beta_{1}$?

Alexis
  • 26,219
  • 5
  • 78
  • 131
CC25
  • 33
  • 4
  • write the likelihood of your data – gunes Feb 11 '20 at 00:01
  • You don't need any assumptions about the $\varepsilon_i$ except that they have zero expectation. Thus, a likelihood-based derivation would not clearly reveal the underlying concepts, which are brought out most strongly in the [algebraic formula of the "normal equations"](https://stats.stackexchange.com/questions/54943) and [the geometric perspective.](https://stats.stackexchange.com/a/113207/919) See https://stats.stackexchange.com/a/46508/919 (*inter alia*) for details and working code. – whuber Feb 11 '20 at 17:12

0 Answers0