Optional Section: Continuous Probability Distributions

Transcription

1 6 Optional Setion: Continuous Probability Distributions 6.5 The Normal Approximation to the Binomial Distribution For eah retangle, the width is 1 and the height is equal to the probability assoiated with a speifi value. Therefore, the area of eah retangle (width 3 height) also represents probability. The normal distribution has many appliations and is used extensively in statistial inferene. In addition, many distributions, or populations, are approximately normal. The entral limit theorem, whih will be introdued in Setion 7.2, provides some theoretial justifiation for this empirial evidene. In these ases, the normal distribution may be used to ompute approximate probabilities. Reall that the binomial distribution was defined in Setion 5.4. Although it seems strange, under ertain irumstanes, a (ontinuous) normal distribution an be used to approximate a (disrete) binomial distribution. Suppose X is a binomial random variable with n 5 10 and p The probability histogram for this distribution is shown in Figure Reall that in this ase, the area of eah retangle represents the probability assoiated with a speifi value of the random variable. For example, in Figure 6.84, the total area of the retangular shaded regions is P(4 # X # 7). p(x) p(x) Figure 6.83 Probability histogram assoiated with a binomial random variable defined by n 5 10 and p Figure 6.84 The area of the shaded retangles is P(4 # X # 7). The probability histogram orresponding to the binomial distribution in Figure 6.83 is approximately bell-shaped. Consider a normal random variable with mean and variane from the orresponding binomial distribution. That is, m 5 np 5 (10)(0.5) 5 5 and s 2 5 np(1 2 p) 5 (10)(0.5)(0.5) The graph of the probability density funtion for this normal random variable, along with the probability histogram for the original binomial random variable, are shown in Figure Figure 6.85 Probability histogram and probability density funtion. The graph of the probability density funtion appears to pass through the middle top of eah retangle and is a smooth, ontinuous approximation to the probability histogram. 6-1

2 6-2 Chapter 6 Optional Setion: Continuous Probability Distributions Remember that for a normal random variable, probability is the area under the graph of the density funtion. In this ase, the total area under the probability density funtion seems to be pretty lose to the total area of the probability histogram retangles. Therefore, it seems reasonable to use the normal random variable to approximate probabilities assoiated with the binomial random variable. There is still one minor issue, but it is easily resolved. Here is an example to illustrate the approximation proess. Example 6.12 Suppose X is a binomial random variable with 10 trials and probability of suess 0.5, that is, X, B(10, 0.5). a. Find the exat probability, P(4 # X # 7). b. Use the normal approximation to find P(4 # X # 7). Solution a. Use the tehniques introdued in Setion 5.4. P(4 # X # 7) 5 P(X # 7) 2 P(X # 3) Convert to umulative probability Use Table I in the Appendix. The symbol, d means is approximately distributed as. The width of eah retangle is 1, so the edges of a retangle above x are x and x For example, the retangle above 4 has edges 3.5 and Simplify. b. For X, B(10, 0.5), m 5 5 and s Figure 6.85 suggests that X, d N(5, 2.5). Using this approximation, P(4 # X # 7) < P(4 # X # 7) Use the normal approximation. 5 Pa 4 2 5!2.5 # X 2 5!2.5 # 7 2 5!2.5 b Standardize. 5 P(20.63 # Z # 1.26) Use Equation 6.8; simplify. 5 P(Z # 1.26) 2 P(Z # 20.63) Use umulative probability Use Table III in the Appendix This approximation isn t very lose to the exat probability, and Figure 6.85 suggests a reason. By finding the area under the normal probability density funtion from 4 to 7, we have left out half of the area of the retangle above 4 and half of the area of the retangle above 7. To apture the entire area of the retangle above 4 and 7, we find the area under the normal probability density funtion from 3.5 to 7.5. This is alled the ontinuity orretion. We orret a ontinuous probability approximation assoiated with a disrete random variable (binomial) by adding or subtrating 0.5 for eah endpoint, to apture all the area of the orresponding retangles. Figures 6.86 and 6.87 illustrate the ontinuity orretion in this example. Figure 6.86 This approximation exludes half of the area of the retangles above 4 and 7. Figure 6.87 This approximation inludes more of the area/probability assoiated with 4 and 7.

3 6.5 The Normal Approximation to the Binomial Distribution 6-3. Using the ontinuity orretion, P(4 # X # 7) < P(3.5 # X # 7.5) Use the normal approximation with a ontinuity orretion. 5 Pa !2.5 # X 2 5!2.5 # !2.5 b Standardize. 5 P(20.95 # Z # 1.58) Use Equation 6.8; simplify. 5 P(Z # 1.58) 2 P(Z # 20.95) Use umulative probability Use Table III in the Appendix Simplify. This probability found using the ontinuity orretion, , is muh loser to the true probability than the omputed value in part (b). Figures 6.88 and 6.89 show tehnology solutions. Figure 6.88 For X, B(10, 0.5), P(4 # X # 7) using umulative probability. Figure 6.89 The normal approximation without and with the ontinuity orretion. Note: Remember that a round-off error is introdued in the above alulations. The alulator results shown in Figures 6.88 and 6.89 are more aurate and better illustrate the impat of the ontinuity orretion. The previous example suggests the following approximation proedure. The Normal Approximation to the Binomial Distribution Suppose X is a binomial random variable with n trials and probability of suess p, X, B(n, p). If n is large and both np $ 10 and n(1 2 p) $ 10, then the random variable X is approximately normal with mean m 5 np and variane s 2 5 np(1 2 p), X, d N(np, np(1 2 p)). A Closer L ok 1. There is no threshold value for n. However, a large sample isn t enough for approximate normality. The two produts np and n(1 2 p) must both be greater than or equal to 10. This is alled the nonskewness riterion. If both inequalities are satisfied, then it is reasonable to assume that the distribution of the approximate normal distribution is symmetri, entered far enough away from 0 or 1, and with the values m 6 3s inside the interval 30, When using the ontinuity orretion, simply think arefully about whether to inlude or exlude the area of eah retangle assoiated with the binomial probability histogram. Consider a figure to illustrate the probability question. This will help you deide whether to add or subtrat 0.5.

4 6-4 Chapter 6 Optional Setion: Continuous Probability Distributions 3. For n very large, for example, n $ 5000, probability alulations involving the exat binomial distribution an be very long and inaurate, even using tehnology. In this ase, the normal approximation to the binomial distribution provides a good, quik alternative. Example 6.13 The World Cup The 2014 FIFA World Cup ompetition was held in Brazil. Despite being a soerenthusiasti nation, a poll prior to the ompetition showed that only 48% of Brazilians supported hosting the event. 33 The support atually dropped sine 2008, perhaps beause of ost overruns, and onstrution delays and aidents. Suppose 500 Brazilians are seleted at random. Use the normal approximation to the binomial distribution to answer the following questions. a. Find the probability that at least 255 of the Brazilians seleted support hosting the World Cup. b. Find the probability that exatly 260 of the Brazilians seleted support hosting the World Cup.. Suppose 215 Brazilians seleted favor hosting the World Cup. Is there any evidene to suggest that the proportion who favor hosting the World Cup has dropped (from 0.48)? Justify your answer. Solution a. Let X be the number of Brazilians seleted who favor hosting the World Cup. X is a binomial random variable with n and p : X, B(500, 0.48). Use Equation 5.8, page 218, to find the mean and variane. m 5 np 5 (500)(0.48) s 2 5 np(1 2 p) 5 (500)(0.48)(0.52) For n and p , hek the nonskewness riterion. np 5 (500)(0.48) $ 10 and n(1 2 p) 5 (500)(0.52) $ 10 Both inequalities are satisfied. The distribution of X is approximately normal with m and s : X, d N(240, 124.8). The probability that at least 255 Brazilians favor hosting the World Cup is P(X $ 255) < P(X $ 254.5) 5 Pa X $ b!124.8!124.8 Standardize P(Z # 1.30) Simplify, use umulative probability Use Table III in the Appendix Simplify. Use the normal approximation with a ontinuity orretion. The probability that at least 255 Brazilians seleted favor hosting the World Cup is approximately Figure 6.90 shows part of the graph of the binomial probability histogram and approximate normal probability density funtion. We need to inlude Figure 6.90 Using the normal approximation, start at x

5 6.5 The Normal Approximation to the Binomial Distribution 6-5 Reall that for any ontinuous random variable X, the probability X equals a single value is 0. Therefore, we must use a ontinuity orretion to approximate the area (probability) of a retangle orresponding to a single value. the area of the retangle orresponding to 255. Using the ontinuity orretion with the normal approximation, start at b. The probability that exatly 260 Brazilians favor hosting the World Cup is P(X 5 260) < P(259.5 # X # 260.5) P a # X 2 240!124.8! # b!124.8 Use the normal approximation with a ontinuity orretion. Standardize. 5 P(1.75 # Z # 1.84) Use Equation 6.8, simplify. 5 P(Z # 1.84) 2 P(Z # 1.75) Use umulative probability Use Table III in the Appendix Simplify. The probability that exatly 260 Brazilians seleted favor hosting the World Cup is approximately 72. Figure 6.91 shows part of the graph of the binomial probability histogram and approximate normal probability density funtion. To approximate the area of the single retangle orresponding to 260, use the endpoints of the retangle Figure 6.91 Use the ontinuity orretion to approximate the area of a single retangle x. The laim made following the poll is that p This implies that the random variable X has a binomial distribution with n and p Using the normal approximation to the binomial distribution, X has approximately a normal distribution with m 5 (500)(0.48) and s 2 5 (500)(0.48)(0.52) Claim: p m X, d N(240, 124.8) The experimental outome is that 215 Brazilians seleted favor hosting the World Cup. Experiment: x Beause x is to the left of the mean, and beause we are searhing for evidene that the proportion of Brazilians who favor hosting the World Cup has dropped, we will onsider a left-tail probability. Likelihood: P(X # 215) < P(X # 215.5) 5 P a X # b "124.8 "124.8 Standardize. 5 P(Z # 22.19) Use Equation 6.8; simplify Use Table III in the Appendix. Use the normal approximation with a ontinuity orretion. Conlusion: Beause the tail probability is so small (less than ), x is a very unusual observation (if p ). Therefore, there is evidene to suggest that the laim is wrong, that p, 0.48, that is, that the proportion of Brazilians who favor hosting the World Cup has dropped.

6 6-6 Chapter 6 Optional Setion: Continuous Probability Distributions Figure 6.92 shows a tehnology solution. Figure 6.92 Approximate probabilities using the normal distribution. Setion 6.5 Exerises Conept Chek True/False The normal approximation to the binomial distribution an be used for any values of n and p True/False Using the normal approximation to a binomial distribution, the probability of a single value is always True/False To use the normal approximation to the binomial distribution, n must be at least Short Answer If X, B(n, p), find the mean and variane of the approximate normal random variable Short Answer Explain why using the ontinuity orretion provides a better approximation than not using the ontinuity orretion. Pratie Suppose X is a binomial random variable with n trials and probability of suess p. In eah problem below, hek the nonskewness riterion and find the distribution of the orresponding approximate normal random variable. a. n 5 30, p b. n 5 85, p n 5 340, p d. n 5 605, p Suppose X is a binomial random variable with n 5 25 trials and probability of suess p a. Chek the nonskewness riterion. b. Find the distribution of the orresponding approximate normal random variable.. Carefully sketh the probability histogram for the binomial random variable and the probability density funtion for the normal random variable on the same oordinate axes Suppose X is a binomial random variable with n trials and probability of suess p a. Chek the nonskewness riterion. b. Compute the values m 6 3s.. Find the distribution of the orresponding approximate normal random variable. Using the results from parts (a) and (b), is the normal distribution a good approximation to the binomial distribution? Justify your answer. d. Carefully sketh a portion of the probability histogram for the binomial random variable and the probability density funtion for the normal random variable near 0 on the same oordinate axes. Explain geometrially why the approximation is or is not aurate Suppose X is a binomial random variable with n 5 25 trials and probability of suess p Find eah of the following probabilities using the binomial random variable, the approximate normal random variable without the ontinuity orretion, and the approximate normal random variable with the ontinuity orretion. a. P(X # 14) b. P(X, 14). P(16 # X # 19) d. P(X. 10) Suppose X is a binomial random variable with n trials and probability of suess p Find eah of the following probabilities using the normal approximation to the binomial distribution with a ontinuity orretion. a. P(X. 220) b. P(X # 198). P(190, X, 200) d. P(X 5 212) Appliations Fuel Consumption and Cars Digital Radio UK reently reported that 55% of all new ars sold in the United Kingdom are equipped with a digital radio. 34 Suppose a random sample of 100 UK new ar purhases is obtained. Answer eah question using the normal a. Find the approximate probability that at least 60 new ars are equipped with a digital radio. b. Find the approximate probability that fewer than 57 new ars are equipped with a digital radio.. Find the approximate probability that between 45 and 55 (inlusive) new ars are equipped with a digital radio Tehnology and the Internet Many ompanies are researhing ways in whih drones ould help improve their business. For example, Amazon ould use drones to deliver

7 6.5 The Normal Approximation to the Binomial Distribution 6-7 pakages more quikly and pizza ould be delivered to your door faster and hotter. Despite the many possibilities, results from a Pew Researh Center poll indiated that 63% of Amerians believe that personal and ommerial drones should not be allowed in U.S. airspae. 35 Suppose 160 Amerians are seleted at random. Answer eah problem using the normal a. Find the approximate probability that fewer than 95 Amerians believe drones should not be allowed in U.S. airspae. b. Find the probability that at least 110 Amerians believe drones should not be allowed in U.S. airspae.. Find the approximate probability that exatly 97 Amerians believe drones should not be allowed in U.S. airspae Eduation and Child Development In May 2014, the National Center for Eduation Statistis (NCES) released results from the National Assessment of Eduational Progress (NAEP). Somewhat disouraging, they suggested that less than 40% of all 12th graders in the United States are aademially prepared for ollege. 36 Suppose 200 U.S. 12th graders are seleted at random. Answer eah problem using the normal a. Find the approximate probability that at least 80 12th graders are prepared for ollege. b. Find the approximate probability that between 90 and 100 (inlusive) 12th graders are prepared for ollege.. Suppose the NAEP results for eah student are used to find that 68 (of the 200) students are prepared for ollege. Is there any evidene to suggest that fewer than 40% of 12th graders are prepared for ollege? Justify your answer Biology and Environmental Siene There have been many siene onferenes and researh papers onerning global warming. Several limate model preditions inlude higher surfae temperatures, rising sea levels, and larger subtropial deserts. Global leaders ontinue to disuss possible ations to stop or slow these trends. Despite the ominous warnings, a reent survey indiated that only 44% of Amerians said that global warming should be a high priority for politial leaders and governments. 37 Suppose 250 Amerians are seleted at random and eah is asked if global warming should be a priority issue. Answer eah problem using the normal approximation to the binomial distribution. a. Find the approximate probability that at most 110 Amerians believe global warming should be a priority issue. b. Find the approximate probability that fewer than 100 Amerians believe global warming should be a priority issue.. Find the approximate probability that between 115 and 125 (inlusive) believe global warming should be a priority issue Mediine and Clinial Studies When filling a presription, there is often a generi option instead of a brandname drug. The generi drug is usually equally effetive and less expensive. In Canada, 37% of presriptions are filled using brand-name drugs. 38 Suppose 300 presriptions filled in Canada are randomly seleted. Answer eah problem using the normal a. Let X be the number of presriptions filled using a brand-name drug. Find the approximate distribution of X. b. Find the approximate probability that at least 120 presriptions were filled using a brand-name drug.. Suppose 95 of the presriptions were filled using brandname drugs. Is there any evidene to suggest that the laim is wrong, that the true proportion of presriptions filled using brand-name drugs is less than 0.37? Justify your answer Manufaturing and Produt Development There is an ongoing debate about whether to label genetially engineered (GE) foods. In the United States, many proessed foods inlude at least one GE ingredient, for example, GE soybeans, orn, or anola. Suppose 70% of all Amerians believe GE foods should be labeled and a random sample of 500 Amerians is seleted. Answer eah problem using the normal a. Find the approximate probability that exatly 360 Amerians believe GE foods should be labeled. b. Find the approximate probability that at least 340 Amerians believe GE foods should be labeled.. Find the approximate probability that more than 340 Amerians believe GE foods should be labeled Marketing and Consumer Behavior Blak Friday is the day after Thanksgiving and the traditional first day of the Christmas shopping season. Suppose a reent poll suggested that 66% of Blak Friday shoppers are atually buying for themselves. A random sample of 130 Blak Friday shoppers is obtained. Answer eah problem using the normal approximation to the binomial distribution. a. Find the approximate probability that fewer than 76 Blak Friday shoppers are buying for themselves. b. Find the approximate probability that between 77 and 87 (inlusive) Blak Friday shoppers are buying for themselves.. Suppose 90 Blak Friday shoppers are buying for themselves. Is there any evidene to suggest that the laim is wrong, that the true proportion of Blak Friday shoppers buying for themselves is greater than 0.66? Justify your answer Psyhology and Human Behavior We all need to all a ustomer servie representative at some time. It is often diffiult to speak with a live person, and the waiting time an be aggravating. Consequently, 36% of Amerians admit that they have yelled at a ustomer servie representative during the past year. 39 Suppose a random sample of 275 Amerians who reently talked with a ustomer servie representative is obtained. Answer eah problem using the normal approximation to the binomial distribution. a. Find the approximate probability that at least 105 Amerians yelled at a ustomer servie representative. b. Find the approximate probability that between 90 and 100 (inlusive) Amerians yelled at a ustomer servie representative.

8 6-8 Chapter 6 Optional Setion: Continuous Probability Distributions. Suppose 85 Amerians atually yelled at a ustomer servie representative. Is there any evidene to suggest that the laim is wrong, that the true proportion of Amerians who have yelled at a ustomer servie representative is less than 0.36? Justify your answer Stuk on Band-Aids Despite the fat that a majority of aidents happen in one s home, only 44% of Amerians have a first-aid kit in their homes. Suppose a random sample of 300 Amerian homes is obtained. Answer eah problem using the normal a. Find the probability that fewer than 125 homes have first-aid kits. b. Find the probability that more than 145 homes have first-aid kits.. Find the probability that exatly 130 homes have first-aid kits Counting Sheep There are many reasons why some people have a diffiult time falling asleep or staying asleep, for example, too muh affeine, limate, lak of physial ativity, or even a lumpy mattress. In a reent poll, it was reported that for those people who suffer from insomnia, 51% blame stress. Suppose a random sample of 80 adults who suffer from insomnia was obtained and asked if they blame stress. a. The sample size (n 5 80) is relatively small in this example. Why is the normal approximation to the binomial distribution appropriate? b. Find the approximate probability that more than 42 adults blame stress for their insomnia.. Find the approximate probability that at least 38 adults blame stress for their insomnia. d. Find the approximate probability that fewer than 35 or more than 45 adults blame stress for their insomnia. Extended Appliations Where s the Remote? We all misplae the remote ontrol to our television. Although some lost remotes are found in the refrigerator, outside, or in a ar, approximately 50% are stuk between sofa ushions. 40 Suppose a random sample of 150 lost (and found) remotes is obtained. Answer eah problem using the normal a. Find the probability that fewer than 70 lost remotes were found stuk between sofa ushions. b. Suppose at least 80 lost remotes were found stuk between sofa ushions. Find the approximate probability that at least 85 were found stuk between sofa ushions.. Suppose 90 lost remotes were found stuk between sofa ushions. Is there any evidene to suggest that the laim is wrong, that the true proportion of lost remotes found between sofa ushions is greater than 0.50? Justify your answer Eonomis and Finane Online banking is quik and onvenient, but it does present some seurity risks. For those people who have an online bank aount, 47% use their online bank password for at least one other online site. Suppose a random sample of 100 people who have an online bank aount is seleted and asked if they use their bank password for at least one other online site. Answer eah problem using the normal a. Find the approximate probability that fewer than 40 people use their online bank password for at least one other online site. b. Find the approximate probability that at least 50 people use their online bank password for at least one other online site.. Suppose a seond random sample of 100 people who have an online bank aount is seleted and also asked if they use their bank password for at least one other online site. Find the approximate probability that between 42 and 52 (inlusive) people use their online bank password for at least one other online site in both random samples Business and Management In January 2014, Sam s Club announed plans to ut 2% of its workfore. 41 The uts were expeted to affet both managers and hourly employees. Suppose a random sample of 150 Sam s Club employees is obtained. a. Use the binomial distribution to find the exat probability that at most 2 employees will lose their jobs. b. Use the normal approximation to the binomial distribution to find the approximate probability that at most 2 employees will lose their jobs.. Explain why in this ase the normal approximation to the binomial distribution is not very aurate. d. Carefully sketh a graph of the binomial probability histogram and the approximate normal probability density funtion near the mean. Use this graph to justify your answer to part (). e. Suppose 7 of the employees atually lose their job. Is there any evidene to suggest that the Sam s Club laim is wrong, that the true proportion of employees who lose their job is greater than 0.02? Justify your answer using both the binomial distribution and the normal approximation Bazinga Despite sientifi evidene (and the popularity of Sheldon Leonard), 51% of Amerians do not believe in the Big Bang. 42 Suppose 1000 Amerians are seleted at random and asked if they believe in the Big Bang. a. Use the binomial distribution to find the exat probability that 510 Amerians do not believe in the Big Bang. b. Use the normal approximation to the binomial distribution to find the approximate probability that exatly 510 Amerians do not believe in the Big Bang.. Explain why in this ase the normal approximation to the binomial distribution is very aurate. d. Suppose 525 Amerians do not believe in the Big Bang. Is there any evidene to suggest that the proportion of Amerians who do not believe in the Big Bang has inreased? Justify your answer using the normal

9 6.5 The Normal Approximation to the Binomial Distribution 6-9 Challenge Biology and Environmental Siene Reall that a Poisson random variable is often used to model rare events and is a (disrete) ount of the number of times a speifi event ours during a given interval. Suppose X is a Poisson random variable with mean l. Then m 5 l and s 2 5 l. If l. 10, then the random variable X is approximately normal with mean m 5 l and variane s 2 5 l: X, d N(l, l). As l inreases, the approximation beomes more aurate. An appropriate ontinuity orretion should be used whenever this normal approximation to a Poisson distribution is applied. Hops are used in brewing beer and make up about 5% of the total volume but ontribute about 50% of the taste. Suppose farmers spray their hops 14 times per year using a wide variety of pestiides and fungiides. Suppose a hops farmer is seleted at random. a. Use the Poisson distribution to find the probability that the farmer sprays the hops at least 17 times during the year. b. Use the Poisson distribution to find the probability that the farmer sprays the hops fewer than 10 times during the year.. Suppose the farmer sprays the hops between 8 and 20 times (inlusive) during the year. Use the Poisson distribution to find the probability that the farmer sprays the hops at most 12 times during the year. d. Use the normal approximation to the Poisson distribution with the appropriate ontinuity orretion to answer parts (a), (b), and (). e. Suppose the farmer sprays the hops 22 times during the year. In addition to avoiding beer made with these hops, is there any evidene to suggest that the mean number of times the hops are sprayed in a year is greater than 14? Justify your answer.