Here in the definition 5.3.11 on the page 249 they write "The $p$-value associated with a test is the smallest significance level $α$ for which the null hypothesis is rejected."

My question is what is the order here: $\bar{X}>\bar{x}$ or $\bar{X} <\bar{x}$ and why? I.e. when it is the case that $\alpha$ is the smallest significabnce level, if

$H_0:\mu=\mu_0$ versus $H_1:\mu>\mu_0$

OR on the other hand

$H_0:\mu=\mu_0$ versus $H_1:\mu<\mu_0$ ?

  • 1
    $\begingroup$ Is the PDF you link to a legitimate fair use? It would be better in any case to write the complete definition in the question here. $\endgroup$ – David K Oct 30 '18 at 13:06

A p-value less than or equal to $\alpha$ means you reject the null hypothesis and have evidence that the mean reflects the alternate hypothesis. The alternate hypothesis can be either $H_a:\mu>\mu_0, \text{ or } H_a : \mu < \mu_0\text{ or simply } H_a : \mu\neq \mu_0$ depending on the context of the test.

For example, you have not tested the mean of a population for several years so you want to assess whether it has changed. Because you don't know whether it is larger or smaller, it is appropriate to use $H_a : \mu\neq \mu_0$. In another situation, you may suspect the mean has decreased (test scores seem to be lower than normal) in which case you would test using $H_a : \mu < \mu_0$.

Edit: One can never prove an $H_a$ with a single test but only provide evidence to support it. One can never discount the possibility of a type I error. This is the random chance of obtaining a p-value less than $\alpha$. There also exists the possibility of another variable affecting the result. But with more testing and control of other variables, one can be more confident in the result.

For $H_a:\mu>\mu_0$, this is a result in the right tail to the right of a Z score corresponding to $\alpha$ of a normal distribution $\text{ and } H_a : \mu < \mu_0$, in the left tail $\text{ and } H_a : \mu\neq \mu_0$ which can be in either tail.

enter image description here

  • $\begingroup$ Why we cannot prove $H_a$ and can only reject $H_0$? A geometric intuition using the normal distribution curve and at the same time the $p$-value would help. $\endgroup$ – user122424 Oct 30 '18 at 18:33
  • $\begingroup$ See edit for additional information. $\endgroup$ – Phil H Oct 30 '18 at 23:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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