Designing Research Designs For Comparisons Between Treatments
A design is basically a blueprint or detailed description of the arrangement or structure of something or for the effective implementation of an action or procedure, or the end result of this blueprint or detailed description in the shape of a model, product or procedure. The word ‘design’ also refers to the art and science of constructing and designing. The Latin root ‘dietus’ means to devise or make plans. The idea of designing goes back to the earliest history of human beings who had to construct the world they lived in.
Designing in its broad sense is used to describe a wide range of scientific studies and technological artifacts designed to satisfy certain needs of the users. These could be scientific research designs for experiments to test the theories and concepts of science; or experiments to test or verify the effectiveness and efficiency of novel products and methods. Real-life applications may also be developed for various fields such as industrial design (which refers to the design of products and systems for use in industry); architecture (which is the designing part of the process); and computer graphics (a branch in computer science which deals with the representation of physical models and their functions). Designing can also be applied in different contexts.
There are two groups of people involved in designing an experiment: the group that is carrying out the research (the research designers) and the group of people who will be involved in the observations and results of the experiment (the participants or subject matter specialists). Designs are then conceptualized and investigated using the appropriate techniques. The procedures and the strategies needed for design are then formulated and tested. The outcome of this planning and implementation stage of the research design process is the specific outcome of the research.
Designing is useful when it comes to testing theories. When a causal relationship between two variables is well established, then a set of designs can provide a test for this causal relationship. These designs can be used in many situations where there is no good way to test a theoretical claim. However, when the claim is not well-established, then it becomes necessary to explore alternative explanations for the data, and one of these alternative explanations should be considered when designing tests.
An example of good research designs is the comparative study of two items. Two items that have a known relation between their price and quality or quantity, are compared using at least three measurements. Each measurement is then compared with the other item to see if the observed relation is valid. The validity of the relation is determined by conducting a meta-analysis. In a meta-analysis, all the available data about the pairs of items is then analyzed to determine which one was the better choice. A 95% confidence interval is used to indicate the range of values for comparison.
Meta-analyses can also be conducted on smaller samples, like a sample of four fifth graders. In this case, the procedures are almost the same as those of the larger sample. However, because the number of subjects is smaller, valid comparisons can be made between treatment groups without the need for a significant number of comparisons. The use of non-parametric statistical tests can help eliminate bias from the comparisons. Thus, a number of non-parametric tests can be easily introduced into a design, thus increasing the validity of the treatments.