1
$\begingroup$

You toss a coin 163 times and get 50 heads. Test the hypothesis that it is a fair coin, using a two-tailed test and 5% level of significance.

The answer wants each of the following: test statistic, critical value, and accept or reject the hypothesis.

It is quite obvious to me that we will reject the null hypothesis since the 2 z-scores associated with getting less than 50 tails (50.5 using continuity correction) are -4.86 and 4.86, while at a 5% confidence interval, the two z-scores are -1.96 and 1.96, but I am not sure what he wants for the test statistic or critical value, or what those are exactly.

$\endgroup$
  • $\begingroup$ Hint: what distribution does the coin follow? The theorem you want to use afterward is central to your success. $\endgroup$ – Sean Roberson Oct 9 '16 at 5:08
  • $\begingroup$ normal? And wouldn't the critical value be 1.96 since thats where the 2nd deviation is? $\endgroup$ – Remy Oct 9 '16 at 5:15
  • $\begingroup$ binomial actually. I know how to calculate that probability, I'm just not sure what a test statistic or critical value is, exactly. $\endgroup$ – Remy Oct 9 '16 at 5:17
2
$\begingroup$

Since 50 heads out of 163 is a proportion, we should use the relevant one-sample $z$-test (we can do so since 163 is a large number): $$z=\frac{(\hat p-p_0)\sqrt n}{\sqrt{p_0(1-p_0)}}$$ Here $$H_0:p=p_0=\frac12\quad H_1:p\ne\frac12$$ $$\hat p=\frac{50}{163}=0.307\quad n=163$$ so the test statistic is $$z=\frac{(0.307-0.5)\sqrt{163}}{\sqrt{0.5×0.5}}=-4.928$$ The critical values are $\pm1.960$ as you thought. Since $-4.928$ falls outside $[-1.960,1.960]$, we reject the null hypothesis and conclude the coin is biased.

$\endgroup$
  • $\begingroup$ So the test statistic would be -4.928? $\endgroup$ – Remy Oct 9 '16 at 5:21
  • $\begingroup$ @JohnH Yes.${}$ $\endgroup$ – Parcly Taxel Oct 9 '16 at 5:21
  • $\begingroup$ Why isn't it 50-81.5/sqrt(163*.5*.5). They are close to each other, and both nearly impossible, but still. $\endgroup$ – Remy Oct 9 '16 at 5:23
  • $\begingroup$ Oh and just realized you said that that was the test statistic in your original answer, oops. $\endgroup$ – Remy Oct 9 '16 at 5:26
  • $\begingroup$ @JohnH We are testing a proportion that is restricted to $[0,1]$ here, not a continuous value like 50 heads or 17 cars passing by. This difference is subtle and I learned it in my statistics class. $\endgroup$ – Parcly Taxel Oct 9 '16 at 5:27

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.