MCS 140 Homework questions

10. Universities are sometimes ranked for prestige according to the amount of research funding their faculty members are able to obtain from outside sources. Explain why it would not be fair?

(Graded) If absolute dollar amounts are used, then larger universities would likely always be ranked as more prestigious, just by virtue of having more faculty. Instead, use the total dollar amount of grant money divided by total number of faculty, to get a "research dollars per faculty member" measure. This removes university size as a confounding variable.

17. Suppose a researcher would like to determine whether one grade of gasoline produces better gas mileage than another grade. Twenty cars are randomly divided into two groups, with 10 cars receiving one grade and 10 receiving the other. After many trips, average mileage is computed for each car. a. Would it be easier to detect a difference in gas mileage for the two grades if the 20 cars were all the same size, or would it be easier if they covered a wide range of sizes and weights? Explain. b. What would be one disadvantage to using cars that were all the same size?

17a) It would be easier if they were all the same size, as otherwise car size/weight could be a confounding variable. (Other answers are possible here, if they're well argued.) 17b) A disadvantage of using cars that are all the same size is that results will only be applicable to other cars of a similar size. The results won't generalize as broadly as one might like.

19. Suppose the administration at your school wants to know how students feel about a policy banning smoking on campus. Because they can't ask all students, they must rely on a sample. a. Give an example of a sample they could choose that would not be representative of all students. b. Explain how you think they could get a representative sample.

19a) If they only polled the school's swim team, this would not be a representative sample. (Many answers possible here.) 19b) Take the roster for the entire school, and choose students randomly.

1. Suppose that a television network wants to know how daytime television viewers feel about a new soap opera the network is broadcasting. A staff member suggests that just after the show ends they give two phone numbers, one for viewers to call if they like the show and the other to call if they don't. Give two reasons why this method would not produce the desired information. (Hint: The network is interested in all daytime television viewers. Who is likely to be watching just after the show, and who is likely to call in?)

2.1 - First, they would not reach all "daytime viewers"; they would reach only those who were watching the show already, surely a biased sample with regard to the show. Second, those who bothered to call would not even be representative of those who were watching the show. People who want the show to remain on the air would be most likely to call.

*15. Moore and Notz (2014, p. 25) reported the following contradictory evidence: "The advice columnist Ann Landers once asked her readers, 'If you had it to do over again, would you have children?' She received nearly 10,000 responses, almost 70% saying 'No!'... A professional nationwide random sample commissioned by Newsday ... polled 1373 parents and found that 91% would have children again." Using the most relevant one of the Seven Critical Components, explain the contradiction in the two sets of answers.

2.14 - In comparing test results across countries, states, years, etc., we should learn about exactly who was included each time (Component 3), and what additional differences there might be among the groups taking the tests (Component 6). Often, differences in results can be explained by the fact that the groups taking them are different on other measures. In this example we would have learned that we were trying to compare the top 50% of students in one country with the top 3% in another, hardly a fair comparison.

5. Many research organizations give their interviewers an exact script to follow when conducting interviews to measure opinions on controversial issues. Why do you think they do so?

2.5 - So that the interviewers don't change the wording slightly to encourage answers that support their own opinions.

8. Suppose a tobacco company is planning to fund a telephone survey of attitudes about banning smoking in restaurants. In each of the following phases of the survey, should the company disclose who is funding the study? Explain your answer in each case

2.8 - a. Absolutely not. If respondents know who is funding the study, they may try to answer to please the interviewer, who they may assume works for that company. b. Yes. Reports of research results should always reveal the source of their funding. c. No, unless the instructions are very strict about how to conduct the interviews. Even then, the interviewers are likely to try to solicit opinions to please their "employer."

22. Refer to the definitions of sample and population on page a. For each of the following, explain who is in the sample, and what population you think is represented. b. Case Study 1.1, testing the effect of a heartbeat sound in a baby nursery. c. Case Study 1.2, testing the effect of taking aspirin on heart attack rates.

22a) The sample is the collections of mothers and their children that were actually studied. The population here would be all mothers together with their newborn children, or maybe just those in NYC, if we're being more conservative (and if there's any reason to believe the results wouldn't generalize further). 22b) The sample here is the group of 22,071 male physicians studied. And the population might be all male physicians, or maybe all male's in high-stress jobs if we want to be more generous with how broadly these results can be applied. Saying the results apply to all men is a bit of a stretch, and saying they apply to all people generally would not be merited without further study (and more representative samples).

1. Give an example of a survey question that is a. Deliberately biased b. Unintentionally biased c. Unnecessarily complex d. Likely to cause respondents to lie

3.1 - a. "Don't you agree that our whole tax system, which is far too complicated for anyone to understand, should be overhauled?" (The key here is to include a leading question like "Don't you agree that ..." or to include emotional wording like "Because abortion involves killing innocent babies..." b. Questions that are unintentionally biased usually involve ambiguous wording. For example, asking "Do you exercise at least three times a week?" may mean something different to college students than to senior citizens. For instance the former group might not consider riding a bicycle for transportation as exercise, while the latter group does. Terms need to be clearly defined. c. Unnecessarily complex questions often are ones that ask more than one question but allow only one answer, like "Do you support doing away with subsidized school lunches and giving food vouchers to the poor?" d. Questions that require knowledge people aren't likely to have, or questions that have socially desirable or even illegal answers are likely to cause respondents to lie. For example, "How many times in the past year have you used illegal drugs?"

10. Give three versions of a question to determine whether people think smoking should be completely banned in airports. Word the question to be as follows: a. Version 1: As unbiased as possible b.Version 2: Likely to get people to respond that smoking should be forbidden c. Version 3: Likely to get people to respond that smoking should not be forbidden

3.10 - a. Some airports are thinking of completely banning cigarette smoking in the airport. Do you agree or disagree with this policy? b. Don't you agree that because of the known harmful effects of passive smoke, airports should ban all cigarette smoking? c. Don't you agree that in airports when between flights, smokers should have the right to smoke cigarettes, as long as a separate, well-ventilated section of the airport is provided for them?

22. Specify whether each of the following is a categorical or measurement variable. If you think the variable is ambiguous, discuss why. a. Brand of car owned b. Price paid for the last car purchased. c. Type of car owned (subcompact, compact, mid-size, full-size, sports, pickup).

3.22 - a. Categorical. b. Measurement. c. Categorical (and ordinal, but that was not one of the choices given)

25. Specify whether each of the following measurements is discrete or continuous. If you think the measurement is ambiguous, discuss why. a. The number of words in a book. b. The weight of a book c. A person's IQ.

3.25 - a. Discrete. b. Continuous. c. Continuous

8. In February 1998, U.S. President Bill Clinton was under investigation for allegedly having had an extramarital affair. A Gallup Poll asked the following two questions: "Do you think most presidents have or have not had extramarital affairs while they were president?" and then "Would you describe Bill Clinton's faults as worse than most other presidents, or as no worse than most other presidents?" For the first question, 59% said "have had," 33% said "have not," and the remaining 8% had no opinion. For the second question, 24% said "worse," 75% said "no worse," and only 1% had no opinion. Do you think the order of these two questions influenced the results? Explain.

3.8 - The order of the questions most likely influenced the answer to the second one. Since the majority of respondents first acknowledged thinking most presidents have had extramarital affairs, they were then not in a position to describe Clinton's faults as worse than most presidents, when the alleged fault of his was an extramarital affair. (Not that the allegations of perjury and other offenses arose after this poss was taken.)

11. The Sacramento Bee (11 Feb. 2001, p. A20) reported on a Newsweek poll that was based on interviews with 1000 adults, asking questions about a variety of issues. a. What is the margin of error for this poll? b.One of the statements in the news story was "a margin of error of plus or minus three percentage points means that the 43 percent of Americans for and the 48 percent of Americans against oil exploration in Alaska's Arctic National Wildlife Refuge are in a statistical dead heat"

4.11 a. With 1000 adults the margin of error is 1 / 1000 = 0.032, or about 3%. b. Adding and subtracting the margin of error of 3% to each of the two figures given, 43% and 48%, produces intervals that overlap. These intervals are likely to cover the true population values for and against oil exploration, so it's possible that the percentages are very similar. Since these figures are based on the same survey, another way to think about this is to assume that the percent who didn't answer for or against stays the same (at about 9% based on these results). Adding the margin of error to the 43% for oil exploration means that the true percent for oil exploration could be as high as 46%, in which case there would be only 45% against, after accounting for the 9% undecided.

32. Suppose that a gourmet food magazine wants to know how its readers feel about serving beer with various types of food. The magazine sends surveys to 1000 randomly selected readers. Explain which one of the "difficulties and disasters" in sampling the magazine is most likely to face.

4.32 They are most likely to face a low response rate, and those who respond are not likely to be representative of all subscribers. For instance, those who prefer beer with their gourmet meals are probably more likely to respond.

7. Explain whether a survey or a randomized experiment would be most appropriate to find out about each of the following: a. Whether there is a relationship between height and happiness b. Whether a public service advertising campaign has been effective in promoting the use of condoms

4.7 a. A survey, because we only need to measure the two variables and not manipulate people's environment. b. This question could be answered in one of two ways. If the public service campaign had already been conducted, a survey would be appropriate. Ideally, though, a randomized experiment would be done in which various communities were randomly assigned to receive the advertising campaign and were then compared with similar communities that did not receive it.

8. Refer to Case Study 1.1, "Heart or Hypothalamus?" in which Lee Salk exposed newborn infants to the sound of a heartbeat or to silence, and then measured their weight gain at the end of 4 days. a. What is the sample of units for this study? b. What two variables did Salk measure for each unit in the sample? (Hint: He imposed one of these two variables and measured the other one.) Heart beat or silence c. To what population of units do you think the results apply?

4.8 a. The sample of units consists of the infants in the two nurseries. b. He measured whether they listened to the sound of a human heartbeat or silence (this was the measure he imposed on them), and he measured weight gain at the end of 4 days.

1. Explain why it may be preferable to conduct a randomized experiment rather than an observational study to determine the relationship between two variables. Support your argument with an example concerning something of interest to you

5.1 You cannot make a causal conclusion with an observational study, while you often can with a randomized experiment.

13. For each of the following observational studies in the news stories in the Appendix, specify the explanatory and response variables. a. News Story 12: "Working nights may increase breast cancer risk b. News Story 16: "More on TV violence." c. News Story 18: "Heavier babies become smarter adults, study shows."

5.13 a. The explanatory variable is amount of time spent working nights and response variable is whether someone got breast cancer. b. The explanatory variable is how much violent television the participant watched in childhood (1 st and 2 nd grades or 3 rd and 4 th grades) and the response variable is violent behavior 15 years later, in adulthood. c. The explanatory variable is birth weight and the response variable is cognitive ability, measured in a variety of ways, in childhood and adulthood.

19. Is the "experimenter effect" most likely to be present in a double-blind experiment, a single-blind experiment, or an experiment with no blinding? Explain

5.19 The experimenter effect is most likely to be present in an experiment with no blinding, because the experimenter can indicate the favored treatment and the participants can respond to that influence. However, it can also be a problem in a single-blind experiment, because the party who isn't blind to the conditions can try to influence the results.

2. Suppose a study found that people who drive more than 10 miles to work each day have better knowledge of current events, on average, than people who ride a bicycle to work. a. What is the explanatory variable in this study? b. What is the response variable in this study?

5.2 a. From the wording, it follows that the only people included in the study are those who either drive more than 10 miles to work each day and those who ride a bicycle to work. The explanatory variable for each person is which of those methods they use to get to work. b. The response variable is some measure of the person's knowledge of current events. c. Individuals who differ for the explanatory variable (method of getting to work) are also likely to differ on the confounding variable (listening to news on the radio) because those who drive to work are more likely to be listening to new on the radio than are those who ride a bicycle to work. And different values of the confounding variable (listening to news on the radio) are likely to result in different values of the response variable (knowledge of current events). In other words, people who drive a long distance to work have more opportunity to listen to news, and thus are more likely to know current events.

29. Explain which of the "difficulties and disasters" in Section 5.5 is most likely to be a problem in each of the following observational studies, and why

5.29 a. Confounding variables are the problem. Students who take writing courses are probably likely to excel in verbal abilities rather than quantitative abilities, so it is natural that they would score lower on a quantitative exam. b. Remembering the past would be a problem here. The women in social work were likely to have had a larger number of female professors, so no one would stand out in particular. The engineers may have had just one or two female professors, which would make those professors easier to remember.

32. Read the summary of News Story 1 in the Appendix. One of the results reported was that people who participated in the meditation program had better immune system response to a flu vaccine. a. Is a cause-and-effect relationship justified in this situation? Explain. b. Explain the extent to which you think the results can be extended to a population beyond only those who participated in the study.

5.32 a. A cause and effect conclusion is justified because participants were randomly assigned to the meditation or control conditions. b. The participants were volunteers. They were all right-handed and worked for a biotechnology company in Wisconsin. However, the relationship between meditation and immune response would probably not be much different for others, so the sample is likely to represent at least the larger population of people who would be interested in participating in a program like this one.

35. Explain why a randomized experiment allows researchers to draw a causal conclusion, whereas an observational study does not.

5.35 Randomized experiments help rule out the confounding variables as an alternative explanation for an observed relationship. Randomization works to average out those variables over the different groups or treatments.

5. To test the effects of drugs and alcohol use on driving performance, 20 volunteers were each asked to take a driving test under three conditions: sober, after two drinks, and after smoking marijuana. The order in which they drove the three conditions was randomized. An evaluator watched them drive on a test course and rated their accuracy on a scale from 1 to 10, without knowing which condition they were under each time. a. What was the explanatory variable in this experiment? b. What was the response variable in this experiment? c. Was this experiment single-blind, double-blind, or neither? Explain. d. Did this experiment use matched pairs, blocks, or neither? Explain.

5.5 a. The explanatory variable was the condition of being sober, having drunk alcohol or having smoked marijuana. b. The response was driving accuracy. c. Single-blind, because drivers knew the condition but the evaluator was blind to it. d. The same drivers were tested under all conditions, so it was a block design with drivers as blocks.

5. "If you have borderline high blood pressure, taking magnesium supplements may help, Japanese researchers report. Blood pressure fell significantly in subjects who got 400-500 milligrams of magnesium a day for four weeks, but not in those getting a placebo" (USA Weekend, 22-24 May 1998, p. 11) a. Do you think this was a randomized experiment or an observational study? Explain. b. Do you think the relationship found in this study is a causal one, in which taking magnesium actually causes blood pressure to be lowered? Explain.

5a) I think it was a randomized experiment, because most people don't take magnesium unless directed to. 5b) I think so; if it was in fact a randomized experiment, then it would be demonstrating a causal relationship.

8. A psychology department is interested in comparing two methods for teaching introductory psychology. Four hundred students plan to enroll for the course at 10:00 a.m., and another 200 plans to enroll for the course at 4:00 p.m. The registrar will allow the department to teach multiple sections at each time slot and to assign students to any one of the sections taught at the student's desired time. Design a study to compare the two teaching methods. For example, would it be a good idea to use one method on all of the 10:00 sections and the other method on all of the 4:00 sections? Explain your reasoning.

It would not be a good idea to use one teaching method on all the 10:00 am students and the other on all the 4:00 pm students. Any observed differences in that scenario could be attributed to the different meeting times of the lectures. Instead, I'd split the 10:00 am students into two sections, and assign the 10:00 am students randomly to those two sections. Then one teaching method would be used for the first section, and the other method would be used for the second section. Split the 4:00 pm section similarly. Then differences observed between the two sections at 10:00 am can be attributed to the different teaching methods, and nothing else, and similarly for the 4:00 pm sections.

13. Suppose the officials in the city or town where you live would like to ask questions of a "representative sample" of the adult population. Explain some of the characteristics this sample?

It would not be sufficient to include only homeowners...homeowners are likely to be the wealthier segment of the population. You'd exclude those who rent, and homeless people. A representative sample would need to contain the same proportions of men and women as the town population, as well as the same or similar proportions of different racial and ethnic groups. You'd also want the different socioeconomic classes represented in your sample in roughly the same proportions that exist in the town's population. Other factors may bear considering, depending on the questions the town wants answered.

1. Explain why the relationship shown in Table 1.1, concerning the use of aspirin and heart attack rates, can be used as evidence that aspirin actually prevents heart attacks?

This was a randomized experiment. It was also at least single-blinded.