1
$\begingroup$

If one has to suggest, what is it one would call a given object that is to be estimated? We already have a generic term "estimator" on the one hand. When the context is clear, usually there is no need to coin another term to address the target to be estimated. However, for instance, when we are in the context of estimating a given random variable using a set of random variables, e.g. when we are to project a random variable $Y$ on the linear span of a set of random variables $X_{1}, \dots, X_{k}$, then under regular conditions there is exactly one $\beta$ such that $X := (X_{1}, \dots, X_{k})'$ implies $X'\beta$ is the best linear approximation for $Y$ in the mean-square-error sense; in this case one may naturally call $X'\beta$ an estimator for $Y$. My question is, in a general similar situation, how one addresses $Y$? The "estimand"?

$\endgroup$
1
  • $\begingroup$ "Property" is often used in the literature. $\endgroup$
    – whuber
    Commented Oct 20, 2018 at 22:34

1 Answer 1

3
$\begingroup$

The genetic term is estimand (that which is to be estimated).

An estimator is a rule/formula/method for obtaining an estimate of the estimand.

$\endgroup$
1
  • $\begingroup$ Thanks. That was what I was thinking (subject to typo however). The reasoning is simple because it just follows "maximand, summand, integrand, ...". Just want to make sure that usage would not be confusing. $\endgroup$
    – Yes
    Commented Oct 7, 2018 at 15:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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