or narrative captured in audio tape or digital recordings, interview transcripts or field notes; alternatively, text already recorded in minutes of meetings, reports, historical or literary documents, personnel records or newspaper clippings;
images, for example paintings, sketches, drawings, photographic stills, DVD recordings, computer-generated images, posters, advertisements;
numbers that result from the systematic capture of classified, ordered, ranked, counted or calibrated characteristics of a sample or population of cases, for example the number of males and females in an organization or the sizes of supermarkets in square metres of floor space.
Words, phrases, narrative, text and visual images (which are often combined, for example in posters) are usually regarded as ‘qualitative’ data. Data that arise as numbers are ‘quantitative’. What is commonly described as ‘qualitative research’ will usually result in the construction of largely qualitative data, while quantitative research will focus mainly on generating quantitative data, but both types of research will usually be a mixture of both sorts of data. The focus in this text is on the analysis of quantitative data, but Part Three does consider mixed methods and how some forms of qualitative data can be quantified.
Data construction may take place either during the routine capture of information, for example on patients admitted to the accident and emergency department in a hospital, or they may be a result of research activity. Data construction in the latter context will include two key elements: the design of the research, which provides the context within which it is intended to construct data, and the actual capture of the data themselves.
The purpose of any research design is to ensure that the data constructed enable the researcher to address the objectives for which the research was undertaken, for example to answer research questions or to test research hypotheses. Writers of texts on research methods are apt to propose listings of different types of design: for example, there are qualitative designs, quantitative designs and mixed designs (e.g. Creswell, 2009); there are exploratory, descriptive and causal designs (e.g. McGivern, 2009). De Vaus (2002) suggests that all designs in the social sciences fall into one of four main groups: experimental, longitudinal, cross-sectional or case study.
Classifications of different types of research design such as those above imply alternative combinations of elements that are for the most part mutually exclusive. An actual piece of research, however, will usually utilize more than one type of design element. So, any design is usually specific to a particular enquiry and will be a unique combination of elements that involve mixing different types of research in the same project. A design may usefully be seen as a series of ‘sub-designs’, for example a design for the specification and selection of the entities that are to be the focus of the research, a design for the role, construction and measurement of selected characteristics of those entities, a design for the capture of data and proposals for their analysis.
A key element in any research design is the clarification of research objectives. These spell out what the research is designed to show or achieve. The more specific these are, the easier it is to design a piece of research that will construct relevant data and the easier it is to see what kinds of data analysis might be appropriate. Ideally, stated research objectives should consist of two key elements: a statement of the general research area, purpose or aim and more specific research questions or research hypotheses. The general research purpose may broadly be exploratory or verificational, for example it may be ‘to explore, investigate or study the effect of playing background music on consumer behaviour in the retail environment’ or ‘to demonstrate or show that the playing of background music has a significant impact on consumer behaviour in the retail environment’. More specifically, a research question might be ‘What is the effect of playing loud music on the amount spent?’ or, phrased as a hypothesis, ‘The faster the music, the less time customers spend in the retail environment’.
The actual capture of the data will require the use of one or more data capture instruments. For qualitative data, the creation of a record could be by way of manual or electronic notebooks, audio or video recorders, camcorders, still cameras or seeking commentary via open-ended questions in questionnaires, email, web pages, blogs, Facebook, and so on. For quantitative data, the most common way to capture data will be through the use of fixed choice responses in a questionnaire, but these may be of very different types. For example, they may be completed by respondents themselves or by interviewers on behalf of the respondent either in a face-to-face situation or over the telephone. Self-completed questionnaires may be delivered personally, by post or using the Internet. An alternative instrument that is commonly used but seldom explained is the diary. These get respondents to record instances of behaviour as and when they occur and may, for example, relate to records of personal contacts or media use – radio listening, for example, is commonly recorded in this way. Increasingly, however, quantitative data are captured electronically using bar scanners, set meters (for television viewing, for example), passive sensing devices, portable data entry terminals or the Internet.
The data from the alcohol marketing study are largely quantitative and are constructed using an academic cross-sectional survey research design. It was cross-sectional in the sense that the study was treated as a ‘one-off’ with measures taken as a single time period. The alternative would be a longitudinal design where measures are taken at intervals with the express purpose of measuring changes. Although the data used in this book are cross-sectional, in reality they are part of a wider study at the Institute of Social Marketing at the University of Stirling which is a two-wave cohort design, the first study carried out from October 2006 to March 2007 with a follow-up of the same respondents two years later. This nicely makes the point that real designs are combinations of elements. Data capture, too, was a combination of interviewer-completed and self-completed (but personally delivered) questionnaires.
Key points and wider issues
Data do not exist ‘out there’, waiting to be ‘collected’ or ‘discovered’. Rather, they are constructed by individuals within a social, economic, political and moral matrix of possibilities and constraints. They are generated as a result of the human activity of systematic record-keeping using a range of data capture instruments. Data are of very different types. They may be qualitative or quantitative, or some mixture. There are, furthermore, different types of each. Qualitative data may be words, phrases, text, images or a mixture. Quantitative data, as is explained in the next section, may be categorical or calibrated according to some specified metric like years, euros or kilograms. Data come in different qualities (good, adequate and poor), they are all in various ways subject to error and may be judged in different ways, for example in terms of comprehensiveness, accuracy, timeliness, relevance, and so on, so what counts as ‘good’ data may not, in any case, be clear-cut.
Data construction may be a routine process or it may be a result of research activity, in which case the construction entails the design of the research and the purposeful capture of the data. A research design is specific to a ‘piece’ of research and is made up of a number of elements whose combination is usually unique to that research.
The wider implications of this view of the nature of data for data analysis are that, before any analysis takes place, the researcher should think about the quality of the data, the design of the research and the contexts in which the data were captured. There is little point, for example, in fussing over the finer points of statistical analysis if the data are of dubious quality to begin with or are inappropriate for the purposes of achieving the research objectives.
The structure of quantitative data
All quantitative data result from the systematic capture of classified, ordered, ranked, counted or calibrated characteristics of a sample or population of entities like individuals, groups, organizations, societies, nation-states, places or objects. They have a structure that consists of three elements: cases, properties and values. Cases are the entities that are the focus of inquiry; they are instances of conditions, behaviours, events or circumstances whose characteristics merit the researcher’s attention; properties are the characteristics of those cases that the researcher has chosen to observe or measure