4
$\begingroup$

Could someone recommend a link or help me out here: where can I find the formula for the regression without an intercept, and how is it deriveed differently than the formula with the intercept? (matrix form)

Improve this question
$\endgroup$
0

1 Answer 1

Reset to default
3
$\begingroup$

It's the same formula you use in regression with an intercept: $\hat{\beta}=\left(X^TX\right)^{-1}X^Ty$. Just make sure there isn't a column of ones for the intercept. It's the same derivation as the formula with an intercept.

Here's a link to some notes that derive regression coefficients: Statistics 512: Applied Linear Models - Chapter 5: Linear Regression in Matrix Form [PDF].

Improve this answer
$\endgroup$
6
  • $\begingroup$ Thank you. If you don't mind me asking another question - if I'm looking to compare variances of the same regression with and without an intercept by looking at their ratio, is there a difference in formulas? $\endgroup$
    – user14386
    Oct 28, 2012 at 20:59
  • 1
    $\begingroup$ Yup. The model without an intercept is actually a special case (nested) of the model with an intercept. Without the intercept, we are actually assuming the intercept is zero. $\endgroup$
    – assumednormal
    Oct 28, 2012 at 21:03
  • $\begingroup$ But if I need to take a ratio of variances, doesn't that mean their formula will just be the same? $\endgroup$
    – user14386
    Oct 28, 2012 at 21:11
  • 1
    $\begingroup$ The residual sums of squares will not be the same unless the regression model with an intercept fits the intercept to zero. $\endgroup$
    – assumednormal
    Oct 28, 2012 at 21:20
  • 1
    $\begingroup$ I apologize. I think I read your initial comment incorrectly. The standard F Test for choosing between two models, one of which is nested in the other, is valid. It's no different because you aren't including an intercept. $\endgroup$
    – assumednormal
    Oct 28, 2012 at 21:22