Chapter 6: Sampling Strategies Review

Match each type of sample with the example that best fits. 1) Cluster sample 2) Stratified sample 3) Systematic sample A) A sociologist first divides registered drivers from a list of motor vehicle registrations into five-year age groups (e.g., 18-22, 23-27, 28-32, etc.), then randomly selects four age groups, and finally randomly selects individuals only from those four age groups B) A sociologist first divides registered drivers from a list of motor vehicle registrations into 5-year age groups (e.g., 18-22, 23-27, 28-32, etc.) and randomly selects 10% of the individuals from each age group C) A sociologist picks every 10th person from a list of motor vehicle registrations to create a sample of registered drivers

1) A 2) B 3) C

A recent Pew poll suggests that Pew is 95% confident that between 65% and 73% of millennials feel negatively about the direction of the U.S. The sample size for this survey was 3,000 millennials. What sample size would Pew need to decrease the margin of error to 1.5%?

12,000 *In probability samples, increasing sample size decreases the margin of error, but there are diminishing returns, meaning that the improvement gained by adding another person to the sample decreases as our sample size gets larger. Margin of error is proportional to the square root of sample size. If Pew wanted to decrease the margin of error from 3 percentage points to 1.5 percentage point (that is, by a factor of 2), they would need to interview four times as many people (because 4 is the square of 2).*

A recent Pew poll suggests that Pew is 95% confident that between 65% and 73% of millennials feel negatively about the direction of the U.S. The sample size for this survey was 3,000 millennials. What is the margin of error?

4% *The margin of error is the amount of uncertainty in an estimate; equal to the distance between the estimate and the boundary of the confidence interval. In this case, the estimate was 69% with a margin of error of 4% with a 95% confidence level.*

Which of the following statements about weighting is true?

A probability sample in which different people have different probabilities of being selected must be weighted for estimates to be accurate *When a sample is weighted, some observations count more than others. The basic principle is this: The more a particular group is overrepresented in the sample, the less weight each individual from that group should receive. For example, if a poll oversampled Asians at twice their proportion in the U.S. population, then each Asian person in the sample would have only half (1/2) the weight of other respondents when researchers calculated estimates for the population as a whole.*

Censuses are likely to become more common due to the rise of what?

Big data *Modern technology, the Internet, and big data have created patterns of data "footprints" everywhere. Rather than using samples, researchers can often gain access to entire databases of transactions, interactions, and other pieces of data.*

Dr. Ortiz wants to carefully follow a select group of college students who are interested in careers with high-profile finance firms. She selects nine students who are currently juniors and seniors and will follow them for a few years to collect rich data on their choices and careers. Which type of approach is this?

Case-oriented * In case-oriented research researchers immerse themselves in an enormous amount of detail about a small number of cases (maybe even only one case). Importantly, choosing a single case to study intensely is still sampling. The researchers need a rationale for why they are studying that case instead of something or someone else and why insights from that case contribute to our understanding of social life. Also, focusing on a single case does not mean that a researcher can learn everything there is to know or talk to everyone who might have something worthwhile to say about it. Even in case-oriented studies, researchers still need to make strategic decisions about what to observe.*

Which of the following statements about censuses is true?

Censuses are more common in social research when the population in question is not composed of people *A census is a study that includes data on every member of a population. Censuses are more common in social research when the population in question is not composed of people. For example, sociologist Edwin Amenta and his colleagues were interested in why some social movements received more coverage in the New York Times than others. To conduct their study, they searched for mentions of an extensive list of social movement organizations in all the issues of the New York Times published in the twentieth century.*

To keep expenses down from having to travel to every state to conduct a survey, a researcher could instead implement which type of sampling strategy in which they would only survey randomly selected individuals in randomly selected states?

Cluster sample *In cluster sampling, the target population is first divided into groups, called clusters. Some of these clusters are selected at random. Then, some individuals are selected at random from within each selected cluster. This sample is still a probability sample because everyone in the population has some probability of being selected.*

Dr. Jackson has designed a survey to examine the effects of mental health resources on the stigma climate on college campuses. He sends his surveys to college administrators and support staff. He finds that greater mental health outreach on college campuses leads to lower levels of stigma on campuses. What is Dr. Jackson's target population?

College campuses *Dr. Jackson is attempting to make generalizations about college campuses using college campus level data (think back to what you learned about units of analysis).*

Why did Eric Klinenberg choose two neighborhoods during the 1995 heat wave in Chicago as his focus for the book Heat Wave?

Contrasting outcomes *Eric Klinenberg wanted to better understand why 750 adults died during the sweltering heat wave that struck Chicago in 1995. He was particularly interested in two adjacent lower-income neighborhoods, North and South Lawndale, that had striking differences in death rates. North Lawndale had 10 times more deaths per capita than South Lawndale, even though both neighborhoods were equally poor.*

Research in which respondents are asked questions when they walk by the corner of an intersection is based on which type of sample?

Convenience *Because the population of people walking at any given intersection is not random, this would be a nonprobability sample, in particular a convenience sample.*

Why did Jerome Karabel choose Harvard, Yale, and Princeton as his focus for the book The Chosen?

Importance *Karabel did not make a list of all elite universities and choose three at random, nor did he choose three schools because he regarded them as most typical. Instead, he selected the "Big Three"--the three schools conventionally regarded as the most elite and most important for developing the political and economic leaders of the next generation: Harvard, Yale, and Princeton.*

Which of the following statements about sampling is true?

It is possible to survey millions of people and get misleading results *Interviewing a sample, or subset, of all Americans will yield accurate results only if we use an appropriate method of selecting people to contact. If we use an inappropriate method, we can ask millions of people and get dramatically misleading results.*

Which method of selection provides random selection into a sample? Selecting based on

Last digit of social security number *Although you might be tempted to say last digit of driver's license, there is still non-random selection into who has a driver's license--poor people and those in urban areas with good public transportation systems are less likely to have a driver's license.*

Which of the following statements about national population-based censuses is true?

Many social scientists believe that the U.S. Census achieves lower-quality results than it would if it were based on sampling *The U.S. Census does not succeed in including everyone in the country. Poor people, urban residents, and people who move frequently--and especially the homeless--are often missed or undercounted. For this reason, many social scientists believe that the U.S. Census achieves lower-quality results than it would if it relied on estimates calculated with a highly sophisticated probability sampling technique.*

When the goal of sampling is to describe a population, there should be ____ between the people who are selected for the sample and the people who are not selected for the sample.

No systematic differences *When the goal of sampling is to describe a population, there should be no systematic differences between the people who are selected for the sample and the people who are not selected for the sample.*

Why did Matthew Desmond choose to study firefighting as his focus for the book On the Fireline?

Past experience and intuition *Matthew Desmond begins his ethnographic study of wildland firefighters by saying, "I began fighting fire because someone asked me to." As a college student, he had taken a summer job as a firefighter. He did not know he would become a sociologist, would return to the firefighting job after starting graduate school, and ended up writing a book about firefighting.*

What are the two advantages of probability samples over nonprobability samples?

Probability samples are unbiased and any difference between the estimates and the true parameter is due to chance *Estimates based on a probability sample are unbiased. That is, to whatever extent estimates differ from the true population parameter, they are equally likely to overestimate it as they are to underestimate it. In a probability sample, the only difference between the estimates and the true parameter is due to chance. This difference is called sampling error. When sampling error alone is responsible for differences between sample estimates and population parameters, we can make precise statements about our uncertainty.*

Which type of sampling strategy is best for case-oriented research?

Purposive sample *Researchers may invest months or years in a case and hence are able to study very few of them. Consequently, case selection is too important to leave to chance. Instead, for the type of research for which case studies work best, purposive sampling will almost always be more informative than choosing cases randomly. In purposive sampling, cases are deliberately selected on the basis of features that distinguish them from other cases. The researchers choose the case on the basis of the features that they believe make the case especially informative.*

Experiments on college students typically allow researchers to detect effects with a ____ than a representative sample of the U.S. population, but require that we test a hypothesis across the range of subgroups.

Smaller sample size *Experiments on college students typically allow researchers to detect effects with a sample size smaller than a representative sample of the U.S. population. If a hypothesis holds true throughout all subgroups of a target population, then we would expect any finding we observe in a representative sample to also hold in whatever nonrepresentative sample we might use. However, we cannot conclude that a hypothesis holds true in the same way throughout all subgroups of a population until we test that hypothesis across the range of subgroups.*

Which of the following is a type of nonprobability sequential sampling strategy?

Snowball sample *A snowball sample is a type of sequential sample in which researchers ask participants in a study to use their social networks to help recruit new participants. If each participant introduces the researcher to a few new participants, the size of the sample can grow--that is, "snowball"--quickly.*

To make sure that smaller groups are represented properly in a study, a researcher might implement which type of sampling strategy to avoid pure chance from causing their sample to be nonrepresentative?

Stratified sample *Stratifying can prevent pure chance from causing samples to be nonrepresentative. Say researchers are drawing a sample of first-year students at a university, and they know that 10% of the students at that university are black. If the researchers want their sample to represent the population as accurately as possible, they could stratify the sample by race and select 10% of the sample from the black group. Using this sampling method, the researchers do not have to worry that they will get an unusual random sample in which the proportion of blacks differs substantially from the proportion of blacks in the university's general population.*

Dr. Olson passes out a survey to his introductory sociology course. He would like to generalize the results of this survey to all college students. Which type of error should Dr. Olson be concerned with?

Systematic error *Systematic error occurs when there are systematic differences between our sample and the target population. In this case, there are almost certainly differences between Dr. Olson's sample (college students in his introductory sociology course) and his target population (all college students).*

In political exit polling, where no sampling frame can be made in advance, which type of sampling makes the most sense?

Systematic sample *The logic of the systematic sample is very important in political exit polls. Researchers do not have a sampling frame--after all, there is no way of knowing in advance who exactly will vote--so they cannot draw a simple random sample in advance. Yet it is extremely important that they use a probability sampling method, because the people who are most eager to tell others how they voted can be a biased sample of all voters. As a result, one common method is that interviewers are instructed to keep track of the order in which individuals leave the polling place and to approach only every nth person about the survey.*

The best sampling strategy depends on ____.

The research question and available resources *The best sampling strategy depends on the research question and available resources, including money, time, and personnel. Researchers use a range of strategies for ensuring that their estimates are as accurate as possible.*

Which statement best describes the sampling error made by the Literary Digest in predicting the 1936 presidential election?

The sample was biased toward people who were higher income *There were two major problems. First, the magazine obtained addresses from lists of people who owned automobiles and telephones, which at that time were still luxuries for many Americans. Roosevelt's support was much stronger among poorer Americans than among wealthier Americans. Second, people who were unhappy with Roosevelt's presidency were likely more motivated to send back their postcards.*

Which of the following is an important potential benefit to nonrepresentative samples?

The similarity of people in a nonrepresentative sample can make it easier to identify a cause-and-effect relationship *The diversity of representative samples makes detecting cause-and-effect relationships more difficult. It is easier to identify a cause when the cases we study are very similar to one another. When cases differ in ways that influence an outcome, we need a larger sample to distinguish the effects of our hypothesized cause, if any, from the effects of these other influences.*