1. Nature Of Sampling

1. Nature of sampling A population is the entire set of persons, objects, events or entities that a researcher intends to study. However, in many instances a population study may not be feasible so the researcher selects a sample or a subset of the target population. Sampling methods refer to the different ways of selecting the sample from the target population. Some sampling methods aim at obtaining a sample that is representative of the target population. Generalising the study results to the rest of the target population can then be made. However, if a sample is biased, generalising from the sample to the population may be less valid or lead to incorrect inferences.

2. Types of sampling methods

•

○

Probability sampling methods

○

Non-probability sampling methods

Probability sampling methods

Each subject or unit in the population has a known non-zero probability of being included in the sample. This allows the application of probability theory to estimate how likely it is that the sample reflects the target population. In statistical terms, a calculation of sampling error can be made. There are several types of probability sampling methods. The first two listed below are frequently used as the sole sampling method or combined with other probability sampling methods.

○

Simple random sampling

○

Systematic sampling

○

Stratified random sampling

○

Multi-stage sampling

○

Cluster sampling

(for details see course handouts) - General advantages

○

A high degree of representativeness is likely

○

The sampling error can be calculated

- General disadvantages

•

○

Expensive

○

Time consuming

○

Relatively complicated

Non-probability sampling methods

The selection of subjects or units is left to the discretion of the researcher and methods are less structured and less strict. Probability theory cannot be used to estimate sampling error. Non-probability sampling methods are usually used for qualitative research when the purpose is exploratory or interpretative. The common types of such sampling methods are:

○

Convenience sampling (or incidental, accidental, chunk)

○

Purposive (or judgmental) sampling

○

Quota sampling (similar to stratified sampling)

○

Snowball sampling

(for details see course handouts) - General advantages

○

Typicality of subjects is aimed for

○

Permits exploration

- General disadvantage

○

Unrepresentative

3.Sample Size Determining what the sample size should be is complex and involves both practical and statistical factors in quantitative research. In qualitative research there is more flexibility and sampling procedures may evolve in the course of a project (see Sarantakos for details).

•

•

Statistical factors to consider:

○

Variability in the population The more heterogeneous the population, the larger the sample needed

○

Desired degree of precision required The greater the degree the precision required, the larger the sample needed.

○

Desired degree of confidence in results If both high precision and high confidence are required the larger the sample needed compared to high precision but lower confidence level, i.e. there is a trade off between precision and confidence.

○

The extent of analysis on subsets of the sample The more analysis carried out on subgroups, the larger the sample needed.

Other factors determining sample size

- Resources available (time, money, personnel) - Expected non-response rates - Expected attrition rates - Expected value of information provided by different size samples compare to their costs. (Factors to consider taken from Diamantopoulos, 1997)

•

Statistical estimation of sample size

The aim of estimating the sample size is to decrease sampling error to a minimum or an acceptable level. Sampling error is related to the standard error. The standard error (SE) is inversely related to the square root of the sample size. For example to halve the standard error, the sample size will need to be quadrupled. To estimate the sample size, an arbitrary size can be decided on, the standard error is then calculated and the sample size increased until the SE is at an acceptable level. Other methods of estimating optimal sample sizes may be used depending on the type of data analysis required. (for details, see Sarantakos 1998)

4. Questions for Reflection 1. How might bias arise in sampling methods? 2. What are possible solutions to reduce this bias?

5. References Diamontapoulos A & Schlegelmilch BB (1997) Taking the fear out of data analysis. London:Dryden. Sarantakos S (1998) Social research. London:Macmillan.

6. Further learning resources The main sampling methods are set out in the Course handouts. Alternatively, you may wish to access the web-based textbook Research Methods Knowledge Base at Cornell University. The address below will link you to the relevant section (Sampling). http://trochim.human.cornell.edu/KB/sampling.htm