C2 Flashcards
how can research be classified?
Nature of research enquiry - exploratory, descriptive, explanatory or causal
Status or source of data - primary / secondary
Type of data - qual / quant
Mode of data collection - ad hoc, continuous
Method of data collection - observation / interview; personal / self-completion; F2F / telephone / online / postal
Type of research design - cross sectional, longitudinal, explanatory or causal, experimental, case study
exploratory research?
undertaken to explore issues or topics. Particularly useful in identifying a problem, clarifying nature of problem or defining issues involved. Can be used to develop propositions & hypotheses for further research
Descriptive?
to answer more clearly defined research questions, aims to build a picture e.g. who is buying brand B / how many people were a victim of hate crime this year. Aims to identify, describe and in some cases quantify things
Causal / explanatory research?
addresses ‘why’ questions, often conclusive e.g. why did people choose brand A over B, why do some prisoners and not others use drugs? These questions allow us to rule out rival explanations and come to a conclusion, to help us develop causal explanations.
Causal explanations
seeks to discover whether one thing (varaible Y) is affected by another thing (variable B)
Covariance and correlation ?
In order to see whether there is a relationship between two variables, you can examine cross tabs of one variable against the other, or plot graphs of one variable against the other by using covariance and correlation statistical techniques.
There may be a direct causal relationship - the change in Y is directly caused by X (e.g. sales and ad spend). But there may be an indirect causal relationship, in the link between Y and X there may be an intervening variable - if you want to rule out possibility that there is another variable involved the research design must allow you to examine its effect.
Important to remember that two things may co-vary, where a change in Y is accompanied by a change in X, and may be due to an extraneous / confounding variable - research should be designed in such a way that you can determine what sort of relationship there is and what is involved.
Inferring causation
You can see covariance, association and correlation but you cannot see causation - it must be inferred. In order to do this you must ensure that research design allows you to do the following:
Look for presence of association, covariance or correlation
If there is a casual relationship between Y & X then it should be expected to see an association between them - a change in Y associated with a change in X. in assessing the evidence for cause you should take into accoun the degree of association between X and Y - you may make one inference on the basis of a strong correlation but a different one on a weak one.
Look for an appropriate time sequence
The effect must follow the cause, thus if X causes Y - X has to precede the effect on Y
Rule out other variables a the cause
A third variable may be the cause of the perceived impact that X has on Y. the ability to rule out other variables rests to some extent on your ability o identify which other variables may be involved.
Come to plausible or common-sense conclusions
Likelihood of explanations, possibility that X may have impacted Y, what other evidence points towards
Primary research
Designed to generate or collect data for a specific problem - data collected does not exist prior to data collection. Can be exploratory, descriptive / causal; qual / quant; syndicated or customised
Secondary research
Process of 2ndry research involves identifying suitable sources, finding sources and gaining access, reviewing them and assessing suitability for your research objectives and evaluating their quality; learning from them; using / assimilating them into own research / thinking about own research or using them to address research objectives.
Secondary research sources
Documents - journal articles and research reports, books
External data - government produced statistics
Data generated by organisation - sales data, data from previous research projects
Quant rsearch?
Involves collecting data from relatively large samples and are typically presented as numbers, often in tables, on graphs and on charts. It provides nomothetic description - general / universal descriptions on a typically large number of case.
Can be collected via census, sample surveys or panels.
Nomothetic?
provides nomothetic description - general / universal descriptions on a typically large number of cases. Qual on the other hand provides idiographic description, that is, description that is rich in detail but limited to few cases.
Idiographic?
It provides nomothetic description - general / universal descriptions on a typically large number of cases. Qual on the other hand provides idiographic description, that is, description that is rich in detail but limited to few cases.
Quant interviews and how they differ to qual?
Quant interviews are structured and standardised - worded exactly the same way and asked in the same order in each interview. Where qual interviews are more like conversations on a spectrum of semi-structured and semi-standardised to unstructured and non-standardised.
Quant interviews can be conducted F2F (in street, central venue, ‘hall test’ / central location test, respondents home or place of work), telephone, post, online.
Uses of quant?
Quant research is useful for describing characteristics of population or market e.g. voting intention, household spending patterns, market and brand share. It is useful for measuring quantifying, validating and testing hypotheses or theories.
Limitations of quant?
Limitations - not flexible - structure and standardisation can produce superficial rather than detailed description and understanding. Closed questions does not allow us to collect answers in respondents own words which can lose ‘real’ responses, detail and context. Can miss out of subtleties in response between respondents how may answer the same thing. These can contribute to low validity.
What is qual?
Typically involves small sample sizes and uses techniques including interviewing, via group discussions, in-depth interviews and workshops, and observations (including ethnography). Concerned with rich and detailed description, understanding and insight rather than measurement. Aims to get below the surface, beyond spontaneous or rational response to the deeper and more emotional response.
Qual limitations?
Limitations - the high flexibility, less structure and standardisation means that this approach is relatively low in reliability, and harder to re-produce at a later date.
Qual represenativeness?
Findings from qual research and not meant to be statistically representative. Sampling strategy should be suited to project objectives in which there is a clear and meaningful relationship between the sample and wider population that it is drawn. The process of selecting sample for qual research should be just as rigorous and systematic as that used in quant.
Qual before quant? and after?
At beginning of study it can be used to generate and develop ideas or hypotheses; to define issues under investigation, to find out how people think, feel and behave, how they tlk about issue or product. This can be useful in structuring quant research and design in qnn.
can also be used post quant study in order to study findings in greater depth and to develop understanding of context.
Qual rsearch uses?
Qual research can be used to:
generate and develop ideas for products, services and advertising
Provide information to guide and develop policy and strategy e.g. marketing, advertising and communications.
Evaluate policies and strategies and their implemenation
Differences between qual and quant? REFER TO TABLE IN GOOGLE DOC
XXXX
what is continuous research? how is it conducted?
Research done on a continuous basis or at regular intervals in order to monitor change over time. Most common method is using panel of respondents chosen to represent target population with data being collected at regular intervals. Panel can be made up of individuals or households, often called a consumer panel, or businesses and other organisation, e.g. retail panels made up of a sample of retail outlets.
Can also use independent samples of the same population, with samples recruited anew for each round of fieldwork. E.g. general household survey or national food survey by UK government.
what is ad hoc research?
Research conducted on a ‘one-off’ basis, to address specific problem or help to understand a particular issue at a certain point of time. E.g. satisfaction of new office accommodation among employees. The types of studies that often come under ad hoc research include advertising pre-tests and communication testings, usage and attitudes studies, hall tests, brand / price trade-off research.
what are observations, why are they used?
Observational techniques based on ethnographic methods are well established in social research and market research. The main advantage of observation over interview is that the responded is recalling or recounting their behaviour to the researcher whereas in observation the researcher sees it first hand - without filter of memory, consciousness or selection. Observation is also useful in following situations:
When do not know / unsure on questions to ask
Starting project in unfamiliar setting
Examining activity or process in a new wau
Observe an individual act in detail
See things happening context
Gather data from another perspective
Greater detail / understanding of a process or behaviour
Observing unconscious or habitual behaviour
Target audience cannot communicate verbally
Concerns about validity or reliability of interview data
Observe behaviour of people en masse
what are quant observations? pros and cons
Quant observation methods tend to be mechanical or electronic. These surveillance methods also tend to unobtrusive, that is, those being observed are largely unaware.they tend to collect data on activity rather than person and activity e.g. traffic counters, devices that record numbers of cars / pedestrians passing a particular point; web counters that count and log visits to a websites. Main advantage of these methods is thoroughness in counting and / or recording activity - main disadvantage is that they can generate high volumes of data that can be difficult to handle, process and/or analyse.
what are qual observations?
Qual observations tend to be done in person by researcher, sometimes with help of camera or voice recorder. These observations tend to be more intrusive, observed are aware that they are being watched and the aim is to collect data on individual and activity - and requires consent of these involved.
What are quant interviews? Types and why you would use them?
Collected via standardised structured or semi-structured ‘forms’ - interview schedules or qnn and diaries. Forms can either be completed by ‘self-completion’ or ‘interview administered’, where the interviewer delivered questions via F2F, telephone and record respondent response within form. Option chosen will depend on:
Study and its objectives e.g. stimulus material may mean that a F2F study may the only feasible one (trying product)
Topics or issues under investigation E.g. subjects of a sensitive nature may be delivered via a F2F interview as it offers a degree of anonymity and distance that a F2F interview does not.
Reaching right sample e.g. telephone or postal may be the only way to reach a hard-to-reach sample
Achieving right numbers e.g. postal survey or email completion rates can be hard to predict
Time and budget available e.g. tight timescales may consider telephone or tight budgets may consider an email surveydue to lack of interviewer cost
what are qual interviews? flexibility?
Where quant interviews are standardised with typically majority closed questions, qual interviews are more like ‘guided conversations’ or ‘conversations with a purpose’ - less structured, standardised, making use of open-ended, non-directive questions
Qual interviews are more flexible than quant, where the interviewer has more freedom to react to what the respondent is saying and adapt the interview accordingly. They can later way questions are asked, order in which questions are asked and can insert follow-up questions if the respondent mentions something the researcher would like to clarify or explore.
Interview over observation?
Choice between interview or observation comes down to two things; nature and objectives or research and practicalities of time and cost. Interviewing may be more suitable when objectives of research are clearly defined and when it is necessary to gather data from a greater range and number of people or settings. Interviewing can be less time consuming than an observational study so it may be cheaper.
Process of research design. Two level system
Process of research design
Research design can be thought of as two levels; the first level is about logic, framework and structure. At this level given what we know about problem to be researched and the sort of research enquiry (exploratory, descriptive or explanatory). The steps within this are:
Defining research problem
Thinking about end use of data
Deciding on which sort of evidence ou need
Deciding on the unit or units of analysis
Deciding logic and structure of research
Choosing research design or structure that will deliver evidence needed
The second level is about deciding mechanics of research - steps below:
Deciding type of data and method of data collection
Primary or secondary or both
Quant / qual or both
F2F, telephone, internet/online, groups, indepths interviews….
Designing a sampling strategy
Identifying the target population
Identifying sampling units and sample elements
Choosing a sampling approach
Choosing a sample size
Designing data collection instrument
Defining concepts, choosing indicators, operationalizing concepts
Designing the questionnaire or discussion guide
Why is research design important
Purpose of research design is to structure research so that it delivers evidence necessary to answer the research problem as accurately, clearly and unequivocally as possible. If the research design is wrong, you will not be able to provide credible claims and what you ‘know’ based on the research. In turn time and money is wasted. The key is to understand the problem to be researched, without a clear idea of what you need to know you will not be able to plan the best way of finding out.
What is validity?
Validity refers to how well a research design (inc research methods and the measures of questions used) delivers accurate, clear and unambiguous evidence with which to answer the research problem. It is an indicator of whether the research measures what it claims to measure.
internal validity?
Internal validity refers to the ability of the research to deliver credible evidence to address research problems. The extent to which the observed results represent truth in the population being studied, and thus are not due to methodological errors.
external validity?
Refers to the extent to which results from study can be applied to the wider population (or setting). Research can not be externally valid unless it is internally valid.
Types of research design (4)
Cross-sectional study
Longitudinal study
Experiment
Study
What is cross-sectional research design?
Cross-sectional design collects data from a cross-section of a population of interest at one point in time. The census of population is an example of cross-sectional design - it describes the make-up of the population at one particular point in time. Ad hoc research is cross-sectional - a single cross-sectional design involves only one wave of data collection, where repeated cross-sectional design involves more than one wave of the research with an independent / fresh sample at each wave. The use of independent sample at each round of data collection is what distinguishes repeated cross-sectional design from longitudinal research. Longitudinal or panel research collects data from the same sample.
uses of x sectional research design
Cross sectional design can be used to provide data to look for and examine relationships between variables; to test out ideas and hypotheses; to help decide which explanation or theory best fits with data; AND TO HELP ESTABLISH CAUSAL DIRECTION BUT NOT TO PROVE CAUSE. Repeated cross sectional can examine trends over time.
With cross-sectional design, and this is what distinguishes it from experimental research design, we rely on there being differences in the sample in order to be able to make comparisons between groups. In experimental research design we create differences within test by manipulating one of the variables - the independent on explanatory variable, in order to see if it causes a change in the dependent variable.
In cross sectional design, having specified the relevant sample and asked relevant questions, we examine the data to see what relationships or differences exist within the sample.
what is longitudinal research design?
Longitudinal research is often referred to as a panel design, with main reason for use being monitoring things, attitudes, perspectives, experiences and behaviour, over time. Longitudinal or panel research is collected from the same sample / panel. Where cross sectional design takes a snapshot of a situation, longitudinal research provides a series of snapshots of the same people over a period of time.
Frequency and number of studies largely depends on research objectives and available budget e.g. if the purpose is for an immediate, short-term impact of an ad campaign a small number of frequent studies may suffice. Longitudinal research is delivered to the sample on more than one occasion, which sets it apart from multiple cross sectional design which takes an independent / fresh sample each time.
Secondary of panel data - in which you re-analyse data for another purpose than the original one, is commonplace.
uses of longitudinal research design?
Main application of longitudinal design is to monitor changes in the marketing or social environment - change that may occur in natural course of things or events that are planned e.g. advertising campaign.
Longitudinal design can be used for descriptive research enquiry. It cannot be used to prove cause, but can be used for:
Exploring / examining relationships between variables
Establishing time order of events / changes and age / historical effects
Help decide which explanation / theory best fits with data
Helps establish causal direction
longitudinal research design - drop out / replacenents? 2 ways of replacing?
As time passes the universal or population from which the panel is recruited changes - this especially true for fast moving markets and in new markets. Panel member themselves also change, as they grow olders. The longer a panel lasts the greater the chance is that panel members will drop out.
If you do not replace panel members you may end up with a small sample, and one that is likely to be unrepresentative (and thus poor in terms of validity), with those who drop out likely being intrinsically different from those who stay.
There are two approaches to replacement:
Find out characteristics of those who drops out and recruit and replace with someone who exactly fits the profile. It is important to note you can never know all of the characteristics of an individual, and that there may be characteristics that do are not used in recruitment but have a bearing on other characteristics nonetheless.
Use of a rolling panel. This is used frequently in any long-term panel which puts a fairly heavy burden on respondents and so results in respondent fatigue and high drop or attrition rates. In this approach panels are refreshed at regular intervals by replacing old or existing members with new ones - ensuring that the overall panel profile remains the same throughout. This smoothes out drop out rates and conditioning (phenomenon of responding to questions in a way that is ‘conditioned’ by having responded to the same or similar questions in previous rounds of data collection).
Data quality reduces with new joiners, as they will have not provided answers data in the same period of those already on period; they are more likely to be enthusiastic (less conditioned) than established panel members and so data may be different and non-comparable with that of established members. To mitigate this data from new panel members should be ignored for the first one or even two data collection periods of their membership.
experimental research design? purpose and how
The purpose of experimental research design allows you to examine the effect of one variable (independent or explanatory variable) in isolation on another (the dependent variable). The idea is that the impact of all other variables will be controlled or removed to clearly see the impact of independent variables alone. This is to determine whether a causal relationship exists and the nature of the relationship, allowing you to rule out effects of all other variables to establish cause and effect. It is the most effective research design in determining causal connections.
Two identical samples or groups are recruited - one is the test group whilst the other is the control group. They are matched on key criteria and the same on all key characteristics. The independent variable (thought to cause or explain change) is manipulated to see effect on the dependent variable, this is referred to as treatment. The treatment is applied to test group but not control group, with any changes to the control group being as a result of factors other than the independent variable. Comparison of these groups allows us to determine the extent of change that is due to the independent variable only. Several post-treatment measurements can be taken as some effects may take longer to manifest or longer term impact of the independent variable may need to be studied.
Experimental design can be impractical and expensive if studying several variables, and impractical if it is needed to determine how sets of variables might interact or work together. To examine the impact of more than one variable factorial design is required. This type of design allows us to examine the main effects of one or two or more independent variables and to look at interaction between them.
experimental research design negatives and mitigations?
Experimental designs are difficult and expensive to use in the real world as it is not always possible to isolate or account for complexity of other variables. Care must be taken in interpreting results, as it is possible that other uncontrolled external factors may act as an influence. It is necessary to be sceptical to extent of causal relationships being proven, even in control groups external factors (known and unknown) may influence one group disproportionally. Also important to think of external validity of the results and the impact of the Hawthorne effect (fact of being studied makes people act differently. You may find that some people drop out, which can be troublesome when you need to take ‘after’ measures of the treatment. Conditioning can be issue in experimental design, where respondents are sensitised to topic and offer post-answers according to pre-treatment answers. Post-stage measure timing is also critical as you do not want to miss effects of the test variables by collecting data too soon or too late, especially too late as people are more likely to drop out.
case study research design?
In-depth investigation of a ‘case’ for research purposes. Case can be a household, organisation, situation, event or individual experience. This research may involve examining all aspects of case, as a whole and its constituent parts. Can be made up of multiple case studies. In designing case studies a framework for the case study that includes all aspects to be examined should be created; research objectives needed to address and uses of this data; sampling strategy; methods of data collection. Variety of data collection methods can be used, including analysis of docs, observation, qual & quant interviewing.
uses of case study rsearch design?
Main application is to get a full picture, to achieve an in-depth understanding and to get detailed (idiographic) description. Useful in understanding the context of attitudes and behaviour to reach greater understanding of their meaning. Case studies are common in educational and organisational research (& evaluation).
If aim is not to understand a particular case fully, and is being used to make generalisations about wider population - care should be taken in ensuring representativeness.
