During my research for a class, I came across a paper that said they estimated an equation using OLS. But the parameter they were estimating appeared to be an expression that looked like this (not the actual equation just something I made up to illustrate):

$$ y = \frac{\beta-1}{2}\chi $$

Rather than what I am used to seeing for OLS:

$$ y = \beta_{0}X_{0} + \beta_{1}X_{1} $$

How does one estimated the equation that way? Or am I missing something?

  • $\begingroup$ Could it be that they estimated the expression in front of $X$ as a whole by OLS and then obtained $\beta$ by multiplying the obtained value by two and adding one? $\endgroup$ – Richard Hardy Feb 25 '15 at 6:58
  • 1
    $\begingroup$ It would be impossible to tell without a reference or a description of what $\chi$ here is. $\endgroup$ – StatsStudent Feb 25 '15 at 7:09
  • $\begingroup$ What was the paper? $\endgroup$ – Glen_b Feb 25 '15 at 10:25

As Richard Hardy said in his comment, in this case you can simply estimate $y=\delta X + \epsilon$. Since $\delta = \frac{\beta - 1}{2}$, you can calculate $\beta = 2\delta + 1$. You do this kind of thing when your regression equation comes from a theoretical model.

Standard errors are simple to calculate in a similar way, using the rules for linear transformations of variances.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.