Why bother to sample in the first place? Surely it must be more accurate to get information from a total group? True, but in terms of practicalities, it will not always be possible to collect information from an entire group. For example, we cannot send a questionnaire to every pregnant woman in Britain as many would have given birth before we found out who should be included. It can also be extremely expensive to gather information from a total group, and it may not always be that much more accurate than a sample. The solution is to select a sample from the population in such a way that the process creates the minimum of bias and represents the characteristics of those in the population as closely as possible. A biased sample would consist of people, events or things that were very different from those in the total group. An example of a biased sample would be a group of pregnant midwives who are asked how they intended feeding their baby. We would expect there to be a difference between this sample and the total population of pregnant women; this would make decisions based on the results unreliable.

It is clear from this that the method of sampling deserves a great deal of thought. We should ensure that it has been planned in such a way as to recognise and minimise potential bias.

Before we select our sample we need to define our target population accurately. This is achieved by specifying inclusion and exclusion criteria. Inclusion criteria are the characteristics we want those in our sample to possess. This is why it is sometimes referred to as eligibility criteria (Polit and Beck 2008). Examples of inclusion criteria would be women who have a normal vaginal birth at term, or women in certain age groups with no complications of pregnancy. In other words, it is the characteristics they must possess to allow them to stand for the general group we want to say something about.

Exclusion criteria consist of those characteristics we do not want those in our sample to possess because it may make them untypical and so bias the results. There may be other reasons for excluding some people from a study, such as the risk of harm for those with a certain medical condition or characteristics.

The researcher must consider the inclusion and exclusion criteria at the planning stage, as these will form part of the detail put into a research proposal or outline of an intended piece of research. These details will also be included in any final report and allow the reader to consider whether the criteria could lead to some limitations in applying the results to other groups. Clear examples of inclusion and exclusion criteria are usually found in randomised control trials (RCTs) that are very sensitive to bias. So, for example, the following appears in the study by McDonald et al. (2010: 90), who looked at the provision of extra support for women following birth to establish influences on length of time breastfeeding:

Different research approaches will require different sampling methods, although some methods can be used in a variety of approaches. In any situation, the researcher must try to draw the sample in such a way as to:

Sampling bias, according to Polit and Beck (2008: 340), is ‘the systematic over-representation or under-representation of some segment of the population in terms of a characteristic relevant to the research question’. Where bias is avoided, or minimised, there is a greater chance that the results can be applied to situations other than the one in which the data were gathered. In other words, it is easier to generalise from the results.

Bias is reduced if the researcher can increase the representativeness of those chosen for the sample. They should match the population they represent as closely as possible in the ways that might influence the outcome of the study. This would include variables such as parity, social class, age and education level. The researcher should establish the distribution of such variables in the population and then demonstrate statistically that the sample does not differ significantly from the total group in the possession of those characteristics.

Table 14.1 outlines the main sampling strategies linked to the various broad research approaches, as there are some clear differences in sampling methods and sample sizes between quantitative and qualitative approaches.

As we saw in Chapter 12, experiments play a key role in establishing the presence of cause-and-effect relationships between variables. To achieve this, sampling must be carried out in very meticulous way so that an accurate conclusion can be deduced from the results. The method of sampling is drawn from a number of options grouped under the heading of random sampling methods. These options form what are called probability sampling methods. Using this approach, every unit in the population, whether it is people, things or events, should have an equal chance of being selected. If this criterion is achieved, it means that some of the more sophisticated statistical tests can be applied to the results. These allow the accuracy of statements made about the results to be calculated. Some of the alternative sampling methods in experimental design are: