• No se han encontrado resultados

Sampling refers to drawing a small subsection of a population in order to study its characteristics (Hair et al., 2016). Population refers to all elements that share a

common characteristics (Hair et al., 2016). Sampling is used when the researcher aims to learn something about a large group by extrapolating from a smaller group

(Easterby-Smith et al., 2018). In certain cases the researchers study all elements of the population – often when the population is not large – which is called a census

(Saunders, Lewis and Thornhill, 2016). However, sampling is crucial if the population is large as otherwise the research would be costly and time-consuming (Hair et al., 2016).

It was also argued that by focusing the resources and attention on studying a smaller subset of the population the researcher may acquire better quality data (Blumberg, Cooper and Schindler, 2014). Additionally, the researchers need to sample in cases where it is impractical to study the whole population (Saunders, Lewis and Thornhill, 2016). For example in cases where the research involves studying product quality of a sealed product and requires opening it for examination (Hair et al., 2016).

Probability sampling

Probability sampling is a sampling technique which is used when every case in the population has a known probability of being chosen for the sample (Sekaran and Bougie, 2016). To create a sample, each case is chosen at random which allows to minimise biases during the selection process (Matthews and Ross, 2010). As a result, the generated sample should be representative of the population (Matthews and Ross, 2010). Thus, allowing to examine a smaller group of respondents and generalise the findings to a larger group (Gray, 2014). This form of sampling tends to be used under a positivist philosophy and predominantly in quantitative research (Collis and Hussey, 2014). Four types of probability sampling are identified – simple random sampling, systematic random sampling, stratified sampling and cluster sampling (Gray, 2014).

For simple random sample, the researcher needs to have a list of the whole population from which a sample of cases is randomly drawn (Easterby-Smith et al., 2018). Each case on the list is assigned a number, the researcher then generates a required number of random values using a random number table or an electronic random

number generating tool and the matching cases are included in the sample (Gray, 2014). It is used when the researcher assumes that population is relatively

homogeneous (Gray, 2014).

Systematic random sampling is similar to simple random sampling, however instead of randomly choosing cases from the list, the researcher picks them at a specific regular interval (Sekaran and Bougie, 2016). To begin the sampling process, the researcher identifies the first case to be included using a random number and then selects subsequent cases at a regular interval (Saunders, Lewis and Thornhill, 2016).

Stratified random sampling involves splitting the population into homogeneous groups called strata and then randomly sampling respondents from each group in order to generate a more representative sample (Hair et al., 2016). This approach is used instead of simple random sample in the cases where the population is not homogenous and there is a chance that a random approach would not generate a representative sample (Easterby-Smith et al., 2018).

Cluster sampling starts by splitting the population into a number of groups or “clusters”, randomly selects a sample of clusters which are then included in the study (Blumberg, Cooper and Schindler, 2014). This approach allows to address the cost implications of other types of probability sampling as the sampled cases could be geographically spread resulting in a high cost and increased time scale of the research (Easterby-Smith et al., 2018).

Non-probability sampling

In contrast to probability sampling, the cases in population of the non-probability sample do not have a known probability of being chosen (Blumberg, Cooper and Schindler, 2014). As a result, the results cannot be directly generalised from sample to the population (Sekaran and Bougie, 2016). Non-probability sampling tends to be used with interpretivist research philosophy as the aim is to describe the environment rather than generalise the results to the overall population (Collis and Hussey, 2014).

Common types of non-probability sampling are – convenience sampling, quota sampling, purposive sampling and snowball sampling.

Convenience sampling involves selecting the cases which are easily available to take part in the study (Easterby-Smith et al., 2018). This approach enables the researcher to efficiently gather large amount of data and pilot test projects (Sekaran and Bougie, 2016). However, it could suffer from selection bias thus not allowing to generalise the results (Hair et al., 2016).

Quota sampling approach splits a population into separate groups and adds cases from each group until a predefined quota is fulfilled (Gray, 2014). Hence, creating a sample which has the same composition as a population (Easterby-Smith et al., 2018).

Quota sampling allows to gather data quickly and from a more varied range of participants than convenience sampling, hence minimising some of the researcher-specific biases (Saunders, Lewis and Thornhill, 2016).

In purposive sampling the researchers select cases due to their usefulness to the research project (Saunders, Lewis and Thornhill, 2016). Thus, the researcher makes a call about who will contribute to the research project. One of the disadvantages is that the researcher may exclude some of the relevant respondents or make some

unconscious biases during selection (Gray, 2014).

Snowball sampling involves initially recruiting a small group of participants from the population who are relevant to the research project and then asking them to suggest further respondents (Bryman and Bell, 2015). This method is useful when there is no information about the population or when the potential participants are hard to reach (Matthews and Ross, 2010).

Sampling of the current project

According to the chosen stances of the current project – internal realism ontology, positivism epistemology, qualitative research method and experiment research – probability sampling would be the most appropriate way of generating a sample (Saunders, Lewis and Thornhill, 2016).

However, within this project it would be time consuming, expensive and therefore not feasible to obtain a list of all pub goers from which a sample could be drawn. Thus, not allowing to use a probability sampling method. Additionally, the first research question of the current project was to examine how venue familiarity influenced visual attention.

Arguably, it is may not even be possible to obtain a population of visitors that are planning to visit a specific pub for the first time, as these decisions are often made because the visitors happen to walk down a street and see a venue. Therefore, the current project used a convenience sample of people with reported intention to

purchase beer intercepted on their entry to a pub. As the current project aimed to study visual attention of consumers in a real pub environment, a group of people who opted to visit a venue are likely to be representative of pub goers. This approach is also consistent with the field-based research location of the current study. As selected sample featured individuals who decided themselves to undertake a shopping journey and the current project measured their actual visual behaviour, it is believed that the

biases associated with the convenience sample are reduced. Additionally, most published research examining the link between visual attention and familiarity, goals and decision-making in the real environment used convenience samples (such as Otterbring et al., 2014; Gidlöf et al., 2017).

Documento similar