3

Since, estimator is a function on a statistical(or any?) data to estimate an unknown parameter, will it be valid (or is one allowed to) to define your own estimator, like your own parameter..

  • 3
    Sure. People are constantly defining new estimators. No reason you can't get in on the fun. – num_39 Mar 29 '23 at 10:49

1 Answers1

7

Sure! All of the common estimators had to be proposed by someone at some point, after all. While there is a great deal of literature that deals with simple and exotic settings, the problems you are solving might be unique, and if you can show why a new or unusual function of the data is a good (in whatever sense you defined “good”) estimator of some facet of your work, go for it.

Among the weirdest estimators I know is that a constant can be considered “admissible” in many (but not all) situations. Yes, this means that there is some defense for going with $\hat \mu = 2$, regardless of the data. This example might give you an idea of how weird you can choose to get when it comes to making your own estimators.

Dave
  • 62,186