Chapter x Sampling



Chapter 8

Sampling

Suggested solutions to questions and exercises

1. What is the difference between the target population and the survey population? Give examples of situations in which the two differ.

The target population is the population from which the results are required; the survey population is the population actually covered by the research. In ideal circumstances, the two should be the same but for practical reasons, they may not be. For example, people or organisations in places that are remote or difficult to access using a face-to-face survey, those on islands, for example, may not be included in a survey population. In a survey of older people's health and social welfare needs, it may be difficult to get permission to interview those living in sheltered or residential accommodation. So although you may have identified them as part of the target population, they may be excluded for the sake of expediency from the survey population.

2. In what circumstances would it be preferable to conduct a census rather than a sample survey?

In situations where the population may be small enough, and accessible enough, for a census to be feasible, for example, research among members of a professional body or employees of an organisation, it may be preferable to do so. In some circumstances it may be important to collect data from all elements of a population, and it may be politic to do so. For example, in research to help with decisions about changes in working practices, it may be advisable to collect data from all employees, in order to demonstrate that the views of all staff are important and will be taken into consideration.

3. Outline the stages involved in developing a sampling plan.

Developing a sampling plan involves the following:

• Defining the target population

• Choosing an appropriate sampling technique

• Finding a suitable sampling frame (if appropriate)

• Deciding on the sample size

• Preparing sampling instructions.

4. What should you consider when deciding between a probability and a non-probability sampling approach?

In deciding whether to use a probability or a non-probability sampling approach, consider the nature and aims of the study:

• If the purpose of the research is exploratory and not conclusive (that is, neither descriptive nor explanatory), in other words, if it is not necessary to obtain highly accurate estimates of population characteristics in order to make inferences about the population, then a non-probability sample is appropriate.

• If it is necessary to obtain measurements from the sample of known accuracy or precision (in order to make statistical inferences or generalisations from the sample to the population), then a probability sampling technique should be used.

It should be borne in mind, however, that probability sampling does not always produce more accurate estimates of population characteristics than non-probability techniques. In fact, in certain circumstances, non-probability (quota) sampling may provide a more representative sample. For example, where there is a little variability within a population, that is, when the population is homogeneous, a non-probability sample can be effective in achieving a representative sample; with a great deal of variability in the population, a random sample is likely to be more effective. When the non-sampling error (errors arising from question wording, interviewer bias, recording error, data processing error) is likely to be greater than the sampling error, non-probability techniques may be just as good at producing a representative sample.

In terms of practicalities, if there is no suitable sampling frame from which to select the sample, then probability sampling is not feasible. If time and budget are restricted, a non-probability sample may be the only option.

5. Describe how you would select each of the following sample types:

(a) a simple random sample

There are two main ways of selecting a simple random sample, the lottery method and the random number method. In the lottery method, each item in the population is represented by a ball/disc. All balls or discs are placed in a container and thoroughly mixed. A sample is then drawn at random. The random number method is similar: each item in the population is numbered from one to N and a sample drawn at random by selecting numbers via a random number generator – a computer or random number table.

(b) a systematic random sample

In selecting a systematic random sample, all items in the population are numbered from one to N and arranged in a random order. Once the sample sample size is agreed a sampling interval is calculated by dividing the population size (N) by the sample size (n). Every N/n th item from the randomised list of the population is selected.

(c) a stratified random sample

It is likely that we know something about the population – for example, the proportion of the population that belong to a particular age group, or the proportion of employees at particular staff grades or the number of organisations in the small and medium sized categories. We can use this information to improve the quality of the sample and the precision of the results by making sure that elements are selected from each stratum. First of all, the population is divided into relevant strata. We then choose the required sample size from each one using simple random or systematic random sampling.

(d) a cluster sample

The population may be clustered in groups – by geography, postal districts, regions, electoral constituencies, for example. We can make use of these clusters in sampling – by sampling from clusters or groups we can cut down on interviewer travel time and so cut down on the cost of fieldwork. Sampling can take place in several stages. First of all, we select a sample of the groups or clusters. These are known as the primary sampling units or PSUs. We then select a sample from within each of these groups or clusters.

(e) a semi-random sample

In a semi-random sample, random starting addresses are selected using a multi-stage stratified random sample. Each interviewer is given a list of random addresses at which to conduct a first interview as well as instructions for choosing which individual to select within that address. He or she is also given instructions for selecting subsequent addresses.

(f) a quota sample.

In designing a quota sample we use our knowledge about the incidence of key characteristics of the population. The aim is to achieve a sample that matches the population on these characteristics. Each interviewer is allocated a quota, detailing the characteristics of respondents needed. The interviewers' task is to select individuals who fit the characteristics in the quota.

6. How would you draw a representative sample for a telephone survey amongst telephone households in the UK?

The first step is to select a representative sample of private households from the telephone directories for the whole of the UK using a fixed sampling interval from a random start point in each area directory. This will give us a representative sample of all households whose numbers are listed in the directory (but it will not give a representative sample of all telephone-owning households because the numbers of some households are not listed). Instead of calling the selected households, this sample is used to create a sample of new numbers by adding one to the final digit of each telephone number. This ensures that the numbers called will include non-listed as well as listed ones. These households should be similar in location and type of household to the number originally selected. However, it under-represents non-listed numbers in areas where non-listed numbers are particularly prevalent.

During the interview, respondents are asked whether his or her number is listed or non-listed, in order to estimate for each directory area the proportion of non-listed households. At the analysis stage, a weight may be applied to each household (if necessary) to correct any differences in the incidence, and therefore the probability of selection, of non-listed numbers within each directory area.

This sampling plan should produce a reasonably representative sample of private telephone households. It is, however, worth noting that:

• it will not represent any mobile telephones, only households; and

• it may over-represent households with two or more land lines since these have a greater chance of being selected (this is a problem with all of the common methods of selecting domestic telephone samples).

Initial calls will be made to the selected ‘Digit plus 1’ numbers during evenings and weekends in order to avoid over-representing households where people are at home during the day. It may be that households with only one line and who are heavy Internet users may be difficult to reach. A practical solution to this is to ensure that subsequent calls are made at different times of day and on different days of the week. (This is managed by our interviewing software, which issues telephone numbers to interviewers in a systematic way.)

7. Why is quota sampling popular in market research? What are its advantages and its limitations?

Quota sampling is popular in market research because in most markets the researcher or the client will have extensive knowledge of the target population, especially on key variables or characteristics. This knowledge will have been derived from primary and secondary sources, including customer databases, geodemographic or national census data and other research. This information can be used to design a sampling framework that will reflect the make-up of the population on these key characteristics.

Quota samples have a number of advantages. In comparison to probability methods, they are relatively quick and inexpensive to set up and administer. Call-backs can be avoided, saving on interviewer travel time and expenses. A quota sample is often a more practicable alternative if a sampling frame does not exist. If a research project, however, demands that results be underpinned by statistical theory, probability sampling is the only choice. While a well-designed probability or random sample should be representative of the target population in all aspects (because of randomness), a well-designed quota sample may only be representative of the population in terms of the characteristics specified in the quota. It may be unrepresentative in other ways. With probability samples, we are able to estimate representativeness; with quota sampling we are not able to estimate representativeness, or even gauge the possible biases that exist.

8. What are the main factors that determine the choice of sample size?

The main factors in determining sample size are:

• Nature and purpose of research

• Importance of decisions to be made on basis of results

• Level of accuracy and precision required of the results

• Type of analysis required

• Budget available

9. Describe the main methods you might use in generating a sample for a qualitative research study, giving examples of the types of studies in which each might be used.

The main methods for generating a sample for qualitative research are:

• Screening

• Convenience sampling – outcropping or lurk and grab

• List sampling

• Network or snowball sampling

• Piggy-backing or multi-purposing

Screening questionnaires can be used to find individuals who match the recruitment or sample criteria. The screening process may be conducted face-to-face by contacting people at home or on the street, or at a specific place where the incidence of those likely to fit the recruitment criteria is relatively high. Recruiting at a specified site is another form of convenience sampling, sometimes known as outcropping or ‘lurk and grab’. For example, if we were looking for church-goers, we would recruit near a church. The recruitment questionnaire may also be administered by telephone or by post. Telephone or postal contact is often used if a sampling frame of the target population, for example, a business directory or database, is available. The client may provide a list of possible contacts or you may have your own list, built up through knowledge of the market or subject area. Using lists or sampling frames in recruitment is sometimes referred to as list sampling. The quality of the sample will depend to some extent on the quality of the sampling frame. Using a network of contacts and asking these contacts to refer you to others is known as network sampling or snowball sampling. This method is useful if a list or sampling frame is not available, or if the sample is difficult to find, for example, a low incidence or low-visibility group. A disadvantage is that you may end up with a sample made up of people with similar characteristics. Piggy-backing or multi-purposing is another way of recruiting or identifying a sample. At the end of a study respondents are asked if they would be willing to be recontacted to take part in further research. They are contacted again at the recruitment stage of the new study to check their willingness to take part and their suitability. This sampling strategy is useful if individuals are expensive or difficult to find.

10. What information would you include about your sampling plan in a report to the client?

• Target population

• Required sample size

• Sampling frame

• Sampling units

• Sample elements

• Sampling technique

• Number of sampling units selected

• Number of sampling units in scope

• Sampling/data collection procedure

• Number of calls

• Response from sample selected

• Sample checks

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download