GROUP TASK #9: Issues in Sampling - DISCUSSION

For each study, identify one or more issues that may potentially cause the sample to be biased.

  1. An orchestra manager is eager to better understand the likes and dislikes of the orchestra’s audience. She distibutes a survey to the first 50 people who arrive at the orchestra’s gala December matinee concert.

    The first people to arrive are likely to differ in several ways from people who arrive later. For example, early arrivals might be especially eager, or especially social, or may come early because they take the bus (perhaps less affluent), etc.

    1. It would be better to distribute survey randomly (using a systematic method like every 10 people).

    In addition, the survey is distributed at only one concert. As a matinee (afternoon) concert, it may attract a different clientele. Also, concerts in December (during the holiday season) may tend to attract a different audience than concerts during the rest of the year.

    1. Better to sample more than one concert.

  2. A researcher is interested in which local radio stations are most listened to by people when driving. The researcher makes a deal with ten local garages. Whenever a car is brought into the shop, the mechanic first turns on the car radio and records the station to which the radio is tuned. After a month of collecting data, the researcher has over two thousand observations.

    In general, this is a pretty good sampling method.

    1. The sample may be biased toward older cars (that require repairs), and/or toward new cars under warrantee.
    2. We have no idea whether a given motorist spends little or lots of time listening to the radio.
    3. The sample may be biased towards white-collar workers and women; blue-collar workers and men are perhaps more likely to do their own repairs. N.B. If the researcher were to ask the participating garages to keep track of the sex of their clients, then the research can compensate for paucity of males. Standing on (several) street corners, the research could estimate the proportion of male-to-female drivers, and so weight the garage results (quota sampling) by sex.

    Notice that it would be bad to use car rental agencies for this sampling method.

  3. A researcher is interested in voice-leading practices. He encodes the complete fugues from the Bach Well-Tempered Clavier and writes software to test various hypotheses in the musical organization.

    1. Biased towards Bach.
    2. Biased towards Baroque practice.
    3. Biased towards fugal practice.
    4. Biased towards voice-leading in keyboard works.

  4. A researcher wants to sample opinions from people attending a nightclub. The quietest places in the club to conduct verbal interviews are the restrooms. Two male and two female research assistants are stationed inside the restrooms where they approach everyone who is about to leave the restroom to answer a couple of quick questions.

    1. The sample may be biased towards those individuals who are drinking (since they are more likely to visit the restroom).
    2. The sample may be biased against those individuals who spend only a little time in the discotheque (“club-hopping”), since they are less likely to visit the restroom.
    3. The sample may be biased towards those individuals who are cooperative/conscientious.
    4. The sample may be biased towards those individuals who are dating a person of the opposite sex (since that person is elsewhere).
    5. There is a slight danger or reinterviewing the same individual.
    6. Depending on the goal, the researcher might want to sample from more than one discotheque. Otherwise, the sampling method looks good. Over several hours, there is a good chance that the researchers will interview the majority of patrons.

  5. Felix Mendelssohn wrote thousands of letters over the course of his life. A historical musicologist conjectures that Mendelssohn became increasingly religious towards the end of his life. From a complete list of letters, the researcher randomly selects 200. Each letter is read by the researcher and coded (yes/no) for whether it contains any spiritual, biblical or religious allusion or content.

    1. Better to use a quota sample in which equal numbers of letters are selected from each decade of life.
    2. Might also want to control for the length of the letters, since longer letters are more likely (by chance) to contain spiritual, biblical or religious content.

  6. A researcher is interested in learning how much musicians earn and their various sources of income. The researcher succeeds in getting the e-mail list for all of the members of the American Federation of Musicians who have provided the AFM with an e-mail contact. An e-mail is sent to members soliciting participation. The e-mail contains a link to the survey’s website.

    1. Biased towards those musicians with an e-mail account.
    2. Biased towards those musicians who are willing to divulge their e-mail account.
    3. Biased towards those who are cooperative/conscientious and will complete the survey. (Biased towards those who have time on their hands.)
    4. Biased towards American and Canadian musicians (AFM serves both countries).
    5. Biased towards unionized (rather than ununionized) musicians.
    6. Biased towards performers (rather than composers, conductors, teachers) etc.

  7. In the early 1980s, Christa Hansen was eager to test whether non-Westerners also show evidence of statistical learning for pitch relationships. Her aim was to recruit listeners who were not familiar with Western musical culture. Along with her husband, Putra, Hansen set out on a small motorcycle toward the remote northeastern region of the island of Bali. Whenever they met someone they asked two questions: “What is the most isolated village in this area? And does it have a gamelan?” Having reached this village, they again asked the same questions, and so continued on their quest. When they were no longer able to use their motorcycle, a footpath took them to a remote village in the shadow of the Gunung Agung volcano. Hansen concluded that she had reached a truly isolated place when the villagers surmised that this fair-skinned stranger must be Javanese. Recruiting twenty-seven participants from this village, Hansen was able to collect data for her experiment.

    A superb sampling approach. Just two problems.

    1. All of her participants were from the same village, so the data aren’t entirely independent from each other. (A few people from several different villages would have been better — although more difficult.)
    2. She collected data from only a single culture, so the conclusions must be described regarding people from the Gunung Agung region.

  8. Sandra Trehub and her colleagues regularly carry out musical studies with newborn infants. From local-area hospitals, they are able to get contact information with names and telephone numbers of mothers who have recently given birth in the various hospitals. They recruit participants by phoning the mothers, describing the basic aims of their experiments, and asking whether they would be willing to participate in an experiment with their new infant. A small sum (roughly $20) is paid to volunteers. Since the research lab is located in a suburban area, there is ample parking. There is also reasonable bus service.

    In general, a pretty good sampling method. Biases:

    1. Western (Canadian) moms.
    2. Biased towards stay-at-home moms with cars.
    3. Biased towards cooperative/conscientious mothers.