5. Choose a Method

Introduction

All that the last posts have covered is developing the background for a study. We've considered what research is, why we need to do research, how to identify potential research questioners, how to look at what others have said about the area, how to refine the question down to objectives and hypotheses, and how to note the assumptions and limitation of research and highlight the significance of a study.

We are now ready to look at the methods you could use to do research. Before looking in more detail at the types of research designs that you may use we will look at the factors which will determine the design you should, or can use; these are the subjects, the variables and the question.

Subjects

To choose an appropriate research design, you need to know how many potential subjects are available and what characteristics your subjects will and won't have.

To summarise what characteristics your subjects will and won't have, you need to write a list of inclusion and exclusion criteria.

Inclusion criteria could include age and sex and set who the population of your study will be. (the criteria are also important for reducing variance, reducing confounding factors)

If there is only one person who meets the inclusion criteria you have set for your subjects, then you know you can only use a single case design. Alternatively, if there are many potential subjects available you can use a group design that has a number of IVs

If the subjects can be randomly selected , you can use an experimental design. Random selection is when every member of the population of interest (that is, who meet the inclusion criteria) has an equal chance of being selected to participate in the study.

Random selection is important in maintaining external validity of a study. It helps you be confident that the results you obtain in your sample are likely to accurately reflect the results you would obtain if you studies all the population, rather than just a sample. That is, the sample adequately represents the population. Many times random selection is not possible and requires being more circumspect in extrapolating results.

Whether you use random selection or not, you should use allocation for experimental designs. If you have a number of groups in your design, such a treatment and control group, you should try to make sure that each of your subjects has an equal chance of being allocated to any one group. This is to help internal validity of the study, by helping the group to be equal to begin with so that any difference can be interpreted as being due to the treatment and not that the groups were different to start with.

Sometimes you may wish to use pre-existing groups, such as Ward A and B. These are called groups of convenience. If you use groups of convenience then you can't do an experimental study, but you can do a quasi-experimental study.

Randomisation is thus important to the validity of your study. Sometimes you may wish to use a pseudo-random selection or allocation. For example, is you wish to keep the number of males and fremales equal you may select and allocate randomly to fill a quota of places for males and females.

In summary, if your subject numbers are very small you may need to use single case designs, and if your subject numbers are large you can use group design. If you can randomly allocate subjects you may be able to use an experimental design.

Variables

Two important characterisitcs about variables which help determine whichi research design is best are control and manipulation.

For an experimental design, the independent variable(s) must be able to be manipulated, that is changed by the investigator. When an IV can;t be manipulated you can use a quasi-experimental, correlation or descriptive design.

The capability to manipulate the IV is a special case of the control an investigator may have over all the variables in the study. Other aspects of control which are important relate to the ability of the researcher to either eliminate or allow for any other variable that could effect the result. We will look at this in more detail with experimental designs.

The strength of an experimental design is the support it can provide for a cause-effect relationship. This is becasue it can control all other potential influences so that the only plausible explanation is that the IV "caused" the difference in the DV. In a quasi-experimental design, because the investigator cannot control all variables, itisi more likely that some other variable could have contributed to any change in the DV. Thus the strength of the support for the casue-effect relationship is much weaker in a quasi-experimental design.

If you are describing a situation there is no need to control the variables.

There is often a trade-off between the amount of control you have in a situation (and so the internal validity) and the closeness of your research to real life (and so external validity). This is particularly important in human and health research because someting that may work in a sterile laboratory may not work in normal life, with all the other physical, mental and social influences interacting on people.

In summary, when you an control all important variables you can do experimental research. when your control is compromised, uyou need to do a quasi-experimental, correlative or descriptive research.

Question

As already mentioned when looking at the different types of research, different questions call for different research designs.

"What is it like?" questions are suited for descriptive studies, "Are they different?" and in particular "Does this cause that?" questions can only be answered with experimental or quasi-experimental studies. Although descriptive studies sometimes use statistics to test of one data set is different to another, this is not experimental as is not answering a cause effect question, just describing what is.