I'm asked to carry out a Z-test for a population proportion on whether a die is fair. The only information I'm given is that "A die fell showing a "four" on $124$ occasions out of $500$".

I'm not sure how to do this, but after giving it a try, this is what I have:

$H_0: p = 0.1667$ (die is fair, outcomes are equally likely)

$H_1: p \neq 0.1667$

Here, I will assume that the significance level is $\alpha = 5\%$

Using the Z-test statistic formula, I have that $Z_{\text{observed}} = 10.9$

Is this right? Would really appreciate some help.

enter image description here

  • 2
    $\begingroup$ Which formula are you using for your calculation. As a hint, you are testing a single proportion against an assumed value. $\endgroup$ – mflo-ByeSE Mar 5 '17 at 22:54
  • 1
    $\begingroup$ @mfloren hi, I added a picture of the formula into the question. Regarding your hint, so do I just do it like a normal Z-test? $\endgroup$ – NDZS Mar 5 '17 at 23:37

If the die is fair, the proportion of $4$'s to expect is $1/6=0.1667$. Is your observed proportion of $p_{\text{ obs}}=124/500=0.248$ far enough from it to exclude the null hypothesis, $H_0: p_{\text{ true}}=0.1667$, based on the observed proportion at a $5\%$ significance level?

You can investigate this in two different ways:

  1. Using asymptotic calculations (normal approximation or Z test):

$$\text{test statistic}= \frac{p_{\text{ obs}} - p_{\text{ true}}}{\sqrt{\frac{p_{\text{ true}}\,(1\,-\,p_{\text{ true}})}{n}}}=\frac{0.248-0.1667}{\sqrt{\frac{0.1667\times(1-0.1667)}{500}}}=4.88$$

  1. Using the binomial test:

$$\Pr(\text{fours} \geq 124)= \sum_{x=124}^{500} {500\choose x} \,0.1667^x\, (1-0.1667)^{500-x}$$

a calculation that explains the reliance on the first method (particularly before modern computers).


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.