How do you do exactly a sample size calculation for a normal distributed population with known standard deviation, to have a chance of type 2 error of at most 5%. .

Zero hypothesis is:

$$ \mu_0 = 20, \sigma = 1$$

Actual mean of the population is known by the researcher:

$$ \mu = 19.5 $$

With the same standard deviation.

Type II error is not-rejecting a false zero hypothesis.

If the chance of a type II error is at most 5%, what is the minimal sample size you have to have for a alpha of 5%?

I would say it is the calculation of:

$$ P(a<Z<b)$$


$$ Z = \frac{\bar x - \mu}{\sigma/\sqrt(n)} $$

I don't know how to progress. Could someone show how it goes?

  • $\begingroup$ $Z$ has a Normal distribution. You have to work out values for $a$ and $b$ based on the test size, and then you can easily find a formula for the power in terms of $n.$ Find the values of $n$ that attain that power (or greater). Where are you stuck in this workflow? $\endgroup$
    – whuber
    Jun 7, 2022 at 17:48
  • $\begingroup$ I don't know the test size, I have to find the sample size. I am stuck $\endgroup$ Jun 7, 2022 at 19:04
  • $\begingroup$ You told us the test size in your question: it's $\alpha=5\%.$ $\endgroup$
    – whuber
    Jun 7, 2022 at 19:30
  • $\begingroup$ Ah I didn't know that was called the test size I thought you referred to number of samples $n$. $\endgroup$ Jun 7, 2022 at 20:04
  • 1
    $\begingroup$ A web search on hypothesis test size would have sorted that out for you; e.g. see en.wikipedia.org/wiki/Size_(statistics) (that definition at the start is slightly inaccurate but works fine for your case); more generally it's given correctly toward the end of this question stats.stackexchange.com/questions/183800/… ... $\endgroup$
    – Glen_b
    Jun 8, 2022 at 2:12


Your Answer

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

Browse other questions tagged or ask your own question.