Study design

Research Design

Research design

Time to spend on this section: 6 hours

It is vitally important to design any kind of research study before you start. Poorly designed studies can give misleading results and having lots of data or information that is not useful, if there is something wrong with the way it was collected.

First you need to know the aims of your study and state them clearly. You may want to estimate the risk of an event, work out how one factor affects another in a system, estimate the value of a parameter for your population of interest, assess people's views on a topic or about an event or evaluate the effect of an action or intervention.

We can split research studies into two broad classes:

Experimental studies: where the researcher deliberately makes some kind of intervention during the study which affects the outcome. These studies include controlled trials where the aim is to vary a single parameter while keeping all the other possible factors under control. Example:

Observational studies: where the researcher does nothing to affect the outcome but observes what happens to a sample of the population of interest. There may still be a control group (e.g. people who happen not to be exposed to a certain factor or risk) but the researcher cannot usually influence its size or membership. Example:

Experimental studies are considered to provide the most convincing evidence in support of a hypothesis because the design allows you to control other factors that may affect the outcome. However, in many cases it is not possible to carry out an experimental study; for instance it may be considered unethical. As an example, a study to investigate the effect on the skin of exposure to ultra-violet light could require a group of patients to be exposed to UV light for different periods of time. As this is known to be carcinogenic it is unlikely that a researcher would get ethical permission to carry out such a trial. Observational studies are normally less controlled and hence are regarded as providing less conclusive outcomes. However it might be ethical to observe a population to see how lifestyle choices that exposed the skin to more or less sunlight affected the incidence of skin cancer and could therefore, provide useful insights for cancer prevention. Observational studies are often used in the absence of a hypothesis either because the work is at an early stage or because it does not lend itself to a hypotohesise-test-refine paradigm.

We can make a second division of study type into:

Longitudinal studies that follow a fixed sample over a period of time. Some longitudinal studies last for many years while others can have a very short time span.

Cross-Sectional studies that are carried out at a fixed point in time. Although repeated cross-sectional studies are sometimes undertaken to assess trends in a population over time, they differ from a longitudinal study in that the membership of the sample is different on different occasions.

You can see the possible combinations of experimental, observational, longitudinal and cross sectional designs in this diagram:

Research Study Designs

Name

Class

Type

Protocol

Examples and uses

Experiment

Experimental

Longitudinal

Select a sample; Measure baseline for parameter, apply intervention, measure outcome. Or divide sample into intervention and control groups. Apply intervention to one group, measure outcome for both and compare.

Before and after measurements such as in a clinical trial for a new therapy, laboratory experiments, field trials e.g. of pesticides or fertilisers.

Cohort

Observational

Longitudinal

Define a cohort and assess current status and influential factors. Observe outcomes.

How a disease progresses,

Whether family income affects educational achievement.

Case-control

Observational

Longitudinal

Define a cohort and divide into those exhibiting and those not exhibiting parameter of interest. Investigate history to assess risk factors.

Whether most lung cancer patients have previously smoked.

Whether parental divorce affects future employment choice.

Cross-sectional

Observational

Cross-sectional

Define a sample and collect information about parameter of interest.

Prevalence estimates, such as how many people have a given disease, or how many people watch a particular television programme.

Repeated cross-sectional

Observational

Cross-sectional

Define a sample and collect information about parameter of interest. Repeat at defined intervals, but not necessarily with the same sample.

Trends, such as how many people are giving up smoking, the effect of publicity on the number of people who cycle regularly, or the popularity of a political party over time.

IDevice IconActivity 1: Considering design

This activity requires you to consider the research that you plan to undertake by answering the following questions. What kind of study will be most appropriate to your research aims? How will the kind of study you use affect the type of outcome you can achieve? What are the limitations of the study method you plan? What are its strengths? Open your log book and write a few paragraphs in the table provided to summarise your thinking on this issue. Download your log book here, if you have not done so already.

Don't forget to save your logbook to your own workspace after you finish or you will loose your work.


In the next three parts you will meet concepts that you need to consider regardless of whether you mean to take a predominantly qualitative or a more quantitative approach to your research.