If $Bias(\hat{\beta}) = (\beta - \mathbb{E}[\hat{\beta}])$, is there a term to describe the quantity $\frac{\mathbb{E}[\hat{\beta}]}{\beta}$?
$\begingroup$
$\endgroup$
3
-
1$\begingroup$ I’ve never seen that, and I think part of the reason is that many parameters are assumed to be $0$ (at least under null hypotheses). Still, when the value is defined, I wonder when this would be useful, so +1. $\endgroup$– DaveOct 26, 2022 at 22:33
-
1$\begingroup$ Variance/scale parameters would be a sensible place for it, and I've seen that sort of summary, but I don't know of any standard name for the ratio. $\endgroup$– Thomas LumleyOct 27, 2022 at 0:51
-
$\begingroup$ @user154510 I've never seen an explicit term for it to my recollection, but in a number of situations it's a perfectly sensible thing to consider. $\endgroup$– Glen_bOct 27, 2022 at 8:55
Add a comment
|