‘Type Il error: Not concluding that the mean lifetime is less than 7 hours when it is. The company will sell batteries that have a shorter lifetime than advertised. (4) Lowering the significance level to 1% would decrease the probability of'a Type | error, but increase the probability of a Type Il error and decrease the power of the test. This would inerease the likelihood of selling batteries with a short lifetime but reduce the number of batteries that are discarded, Test 9B Partt 1. a Area under standard normal curve above = = 1.40 is 0.0808. 2. b_ The question is whether the new product works more quickly, so we are interested in the one-tailed altemative for the mean time for pain relief is less than 25 minutes. 3. e Since the P-value is less than ct, we have sufficient evidence against Ha. to conclude that ‘Apis not likely to be true, Since the testis two-tailed, we can conclude only that p= 60. 4. © rprocedures are robust with respect to the Normality of the population if 7 is large enough, so the most important condition is randomness. Since the null value of 1.5 is in the 95% confidence interval, we cannot reject Ho. b Tis true because power is greater if the aitemative (actual) value of the parameter is farther froin dhe wull, UD is tue because P(Type Lerror) ~ wand P(Type HH error) depends ona, His false, since P(Type | error) =a, not P(Type I error) 7. b Only the Normality of the differences is important, since we are testing the mean of the difference in weights, not the pre-diet and post-diet mean weights 8. © Type Il error is failing to reject Ay when it’s not true, thus concluding that the mean weight loss is zero when it’s actually positive 9 dA test with such a large sample size will ofien produce a significant result for a very stnall departure from the null value. There is little practical significance to a change in finish time of 0.08 seconds 10. © Power is the probability of rejecting a false null hypothesis, which in this case is concluding that the true mean daily Cale um intake for women a the college is lower than 1000 mg when it is actually 900 mg. am Part It TI. (2) State: We wish to test M, : p= 0.67 versus H, : p> 0.67, where p = the true proportion authors who support the system of only publishing longer papers online. We will use a significance level of «= 0.05. Plan: The procedure is a one-sample =-test for a proportion. Conditions: Random: The journal took a random sample of 104 recent authors. Normal: ‘np = (104): (0.67) = 69.7 2 10, and n(1— p)=(104)-(0.33) = 34,3210, Independent: Random selection ensures that individual observations are incepenclent; we will have to assume that there are more than 10104 =1040 recent authors of articles in the British Medical Journal, Do: 452 The Prictice of Satstes, He- Chapter 9 © 2011 BFW Publishers0.69-0.67 0.43; P-value = 0.3336, Conclude: A P-value of 0.3336 is greater than «= 0.05, so we fail to rejact Hl: there is insufficient evidence to conclude that the proportion of recent authors who support the system is greater than 67%. (b) Ifthe true proportion of authors who support the system is 0.67, the probability of getting a sample proportion of authors as far or farther above 0.67 as our sample is 0.3336. 12, (a) Hy :41= 800 versus H, : 42> 800, where = mean concentration of Red Tide toxins in clams, in g/kg. (b) This sample size is too smell for a population that is known to be non- normal. (c) Type I error: Concluding that the mean level of toxin is above 800 pg/kg when it is not. The DMF would close the area to clam harvesting, which would have a negative economic impact on anyone who depends on the clam business, even though the clams are safe to eat ‘Type Il error: Not concluding that the mean level of toxin is above safe levels when it is. This could cause anyone who eats clams from this area to become sick or even die. 4) Raising the significance level to 10% would increase the probability of a Type [ error, but decrease the probability of a Type Il error and increase the power of the test. This would decrease the likelihood of people eating toxic clams, so it might be a good idea, Better safe than sorry. Test 9C 1. € The question is whether the students in t1e Agriculture Schoo! are less ft, which means having a higher mean pulse rate. Thus H should be greater than the null b A Type Il ervor is failing to reject a false #7, which in this context means concluding that the mean pulse of the agriculture students is not above 72 when itis. 3. e £=2.051 is between 0.025 and 0.05 on the table for 27 dearees of freedom, Since the test is two-tailed, we double the P-value, so it is between 0.05 and 0.10. Since the sample size is small, we must mect the condition that the population is approximately Normally distributed (cordition 1 in this question), and itis always important that we can view the sample as an SRS (condition 11). Condition II] is not required for a t-test, which uses the sample standard deviation to estimate 6. A test based on a sample of 4 individuals has very little power, so getting a P-value so close t0 0.05 suggests that the likelihood of a Type Il error is high. A larger sample would increase the power of the test. 6. The P-value measures the probability of getting a result as extreme as the sample statistic is when the null hypothesis is true 7. e Since we already know the salary of every member of this population, we can easily calculate the parameter value, There is no need for statistical inference, © Atthe «= 0.05 level, we expect to havea Type I error—getting a significant result when the null hypothesis is true—5% of the tirie. 1f50 /-tests are performed at this significance tevel and all 50 null hypotheses were true, we would expect, on average, ©2011 BEW Publishers The Practice oF Staisties, 4e- Chapter 9 453