6
$\begingroup$

A beginner's question to check I've understood correctly. A basic stats textbook says:

"The variance of the sampling distribution of the median is greater than that of the sampling distribution of the mean. It follows that sample mean is likely to be closer to the population mean than the sample median. Therefore, the sample mean is a better point estimate of the population mean than the sample median."

Does it follow that for distributions where the median=mean, the sample mean is a better point estimate of the population median than the sample median?

Improve this question
$\endgroup$
3
  • 2
    $\begingroup$ Which text? Where did it say this? Did you leave out anything important, like some prior statement of the distribution for which this claim is made? (I imagine it's probably intended to be under i.i.d. sampling from a normal distribution). As a general claim, it is of course false, so if they don't mention that it only applies in some cases, you should probably have doubts about what else you read in that book. The Laplace is a common counterexample to your question at the end there. $\endgroup$
    – Glen_b
    Commented Mar 18, 2016 at 2:21
  • $\begingroup$ Possible duplicate of For what (symmetric) distributions is sample mean a more efficient estimator than sample median? $\endgroup$
    – Glen_b
    Commented Mar 18, 2016 at 2:27
  • $\begingroup$ I don't think the text of the question matches the title. $\endgroup$
    – David Lane
    Commented Apr 5, 2017 at 0:50

1 Answer 1

Reset to default
3
$\begingroup$

It would depend on details of the distribution family. For normal distributions, what you said would be true. For some more heavy-tailed distribution, it might not. You could for instance check with some t-distribution with low degrees of freedom.

Improve this answer
$\endgroup$

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.