Friday, December 26, 2008

Group Quasi & Experimental Methods

Introduction

This post looks at several group quasi and experimental methods. Remember that this will discuss quasi and experimental methods together, but the level of control, and the strength of generalisations from the two types of methods are different.

Number of Groups

Categorises group quasi/experimental methods by the number of groups in the design. There can be one group of subjects, two groups or more than two (k) groups.

One Group Design

When there is only one group, random allocation is not an issue, but random selection is.

Data can be collected once from a single group and compared to recommended, normative or hypothetical ideal data. Data can be collected twice from the same single group. This is a commonly used design as it allows each subject to be their control, thus reducing the effect of inter-subject differences. Data can be collected k times from the same single group. This has the same advantages of the two data collection desings, but allows one to gather more information. This may be useful to see how long the effect of treatment lasts, look at three treatments and a control, etc..

Two Group Designs

As with one group designs, random selection can be important but allocation to groups is also important. This can be done randomly, or balanced in some way, or matched. Two groups formed by random allocation are called independent groups. Groups formed by matching are called dependent groups. When groups are matched, the researcher has tried to reduce the difference betteen pairs of subjects (thus it is similar to the one group two data collections design and is analysed with the same test(s).

The classic experimental design is a two group design. It involves random allocation (R) a pre-test/pre-observation (O) of each group, a period where treatment (X) is given to one group but not to the other, followed by a posttest/post observation (O) of both groups. In reasearch notation it looks like:

R.......O......X.......O
R.......O...............O

One permutation of the classic design are where a follow-up is added to look at the whether the treatment effects is manitained.

R......O.......X......O......O
R......O...............O......O

Another permutation is to omit the pretest. If randoom allocation is performed, and a large enough sample size is used, teh two groups should be equivalent. Although a pretest is useful to check this, sometimes the testing may have an effect of its own, and so teh researcher may wish to omit it.

R.....X......O
R.............O

K-Group Designs

Investigators sometimes use more than two groups. this allows for more questions to be asked, but makes the design more comlex and so analysis difficult.

A popular 4 group desing is called the Solomon design and was contrived to allow for the effect of an interaction bewteen pretesting and the intervention.

R....O.....X.....O
R....O............O
R...........X.....O
R..................O

Another group of k-group designs are called factorial desings. These are used when the researcher wishes top look at two (or more) IVs in the one study. There can be 2x2 factorial designs with the first IV factor has two levels and the second IV factor has two levels and therefore need 4 (2x2=4)groups. You will see factorial desings looking at 3 factors (IVs) at once and a number of levels, for example a 6x2x3, but the analysis and interpretation gts very difficult.

One of the benefits of this type of study is the ability to look at interactions between IVs (factors).

No comments: