3
$\begingroup$

I'm supposed to perform a hypothesis test with:

sample size = $27$

sample mean = $10.7$

sample standard deviation = $3.6$

Hypothesized mean, $\mu_0=11$

Because the sample size is small and population standard deviation is unknown, I perform a t-test to see if I can reject my null hypothesis.

I find that the t-score is $-0.43$ and:

$\text{p-value} =2p(t_{26} > 0.43)$

This is the t-table I will be provided in tests/exams at my school: https://drive.google.com/file/d/0BxUaD_pbAL2hQzE0dzZtVHhsSlU/view

There is no value close to $0.43$ in the row of $\text{df} = 26$ on the t-table, so what do I do?

Am I supposed to use the z-table instead? Or is using R or some other computational program the only possible way to find the p-value?

Also, regarding the t-table I linked, is this table only for one-tailed t-tests? If so, what do I do if I need to do a two-tailed test (like in this example)?

$\endgroup$
  • $\begingroup$ You have received answers already but I would add that in real life (as opposed to an exam) you would use R or some other program. $\endgroup$ – mdewey Oct 20 '16 at 11:23
2
$\begingroup$

Also, regarding the t-table I linked, is this table only for one-tailed t-tests?

Directly it's for one-tailed but you can use it with two-tailed tests. I'll explain the one-tailed use and then discuss how to do it for two-tailed tests.

There is no value close to 0.43 in the row of df = 26 on the t-table, so what do I do?

That sort of table is able to give bounds on the p-value. That's sufficient to know whether to reject or not.

$$ \begin{array}{r r r r r|r} \hline t_{0.1}&t_{0.05}&t_{0.025}&t_{0.010}& t_{0.005}&\text{df}\\ \hline \vdots&\vdots&\vdots&\vdots&\vdots&\vdots\\ 1.315& 1.706& 2.056& 2.479& 2.779& 26\\ \vdots&\vdots&\vdots&\vdots&\vdots&\vdots\\ \end{array} $$

For example if your t-value was 1.5, which is between 1.315 and 1.706, you know that the one tailed p-value is between 0.1 and 0.05.

With $t=0.433$ you'd only be able to say that the one-tailed p-value was greater than $0.1$ (and for any typical significance level that means you don't reject $H_0$).

what do I do if I need to do a two-tailed test?

Double the one-tailed p-value (or with this table, double the bounds you identify).

In fact the information to double it is already in the formula in your question:

Screenshot showing 2p(t > 0.43)

So if you had $|t|=1.5$ you'd say the two-tailed p-value was between $0.2$ and $0.1$.

If you had $|t|=0.433$ you'd say that the two-tailed p-value was greater than $0.2$.


When your t statistic is between two tabulated values, it is possible to get an approximation to the p-value via interpolation but it's not necessary in this case (and I wouldn't spend the extra time in an exam).

$\endgroup$
  • $\begingroup$ I always wonder why nowadays they still want students to use those tables... $\endgroup$ – Tim Oct 20 '16 at 8:00
  • $\begingroup$ @Tim Because it's cheaper than supplying computers to everyone doing an exam, while struggling to making sure nobody cheats. Some intro stats subjects have huge enrolments. $\endgroup$ – Glen_b Oct 20 '16 at 8:40

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.