Experimental Design#
The key to eliminating bias and making sound statistical inferences is good experimental design.
Definitions#
- Experimental Unit
The entity (person, object, thing) being observed in an experiment.
- Treatment
The condition applied in an experiment.
- Indicator Variable
An indicator variable is the variable over which the researcher has control.
- Response Variable
A response variable is the variable measured by the researcher.
Note
Indicator variables are sometimes known as either explanatory variables or independent variables.
Response variables are sometimes known as either explicated variables or dependent variables.
- Example
A researcher changes the pH concentration of a solution and measures the temperature at various concentrations.
In this example, the indicator or explanatory variable is the pH contentration. The researcher is able to change the amount of acid or base that is added to the solution. The response variable is the temperature of the solution.
Blind Studies#
The concept of blind studies mainly applies to the areas of psychology and medicine. These sciences deal with human nature and humans are, if nothing else, peculiar creatures. Their expectations can influence the results of the experiment.
In medical studies, the test group will receive the real treatment while the control group will receive a placebo treatment. For example, in testing whether a certain drug treats a medical condition, the control group would receive a sugar pill. Yet it sometimes occurs the control group experiences the effects of the real treatment simply by expecting the effect,
Placebo Effect Video
Click on the image to watch the video.
To minimize the Placebo Effect, single blind and double blind studies are performed.
- Single Blind
In a single blind study, the participant does not know whether he or she is in the control or test group. However, the researcher does know.
- Double Blind
In a double blind study, neither the participant nor the researcher administering the treatment know if the individual is in the control or test group.
Matched Pairs#
A matched pairs design is an experimental design where researchers match pairs of participants by relevant characteristics. Then the researchers randomly assign one person from each pair to the treatment group and the other to the control group. This type of experiment is also known as a matching pairs design.
An ideal example of a matched pairs design would be twins,
If one of the twins is submitted to a treatment, their genetically identical counterpart serves as a nearly perfect control.
Randomized Blocks#
A randomized block experiment should be understood a series of identical experiments, where each block of the population sampled is composed of the same distribution of individuals.
A randomized block design is commonly encountered in agricultural applications. Consider a farmer who wants to test a new type of seed against his usual stock to determine if the yield is higher. In a randomized block experiment, he would break his plots of land into blocks, and then partition each block into segments, call them A, B, C and D,
The type of seed would be planted in the A segment of each block, i.e. the A segment would receive a treatment, while the other segments would be planted with the farmer’s usual stock. Data would then be collected from each block and analyzed in isolation to determine if the new type of seed has any benefits.
In essence, each block represents a separate experiment, where the treated group is tested against the control group.