PCE - Types of Experimental Design Lesson
Types of Experimental Design
There are three primary types of experimental design: completely randomized, randomized block, and a special case called matched pairs. In a completely randomized experiment the units or subjects are assigned randomly among all the treatments.
Example: A study is to be conducted to compare the driving distances for 4 different brands of golf balls.
This comparison will involve 40 golfers.
Experimental units: the 40 golfers
Factor: the brand of the golf ball
Treatments: the four brands of golf balls
Response variable: the driving distances
Next the 40 golfers get divided into treatment groups.
GOLFERS are chosen randomly for each group and brand of golf ball is randomly assigned to a group.
The next step would be to compare the driving distances, more specifically the mean of the driving distances, for each brand of golf ball to answer the question "based on this experiment, which ball is the longest in driving?"
In a randomized block experiment the units or subjects are divided into blocks at the beginning of experiment. Experimental units are assigned randomly among all the treatments separately within each block. Side by side experiments are conducted with comparisons and conclusions formed about the blocks separately. There are two conclusions...one for each block.
More about the golfers...
Example: A study is to be conducted to compare the driving distances for 4 different brands of golf balls.
This comparison will involve 40 golfers.
Experimental units—the 40 golfers
Response variable—the driving distances
Factor—the brand of the golf ball
Treatments—the four brands of golf balls
Since driving distances are different for men and women, blocking by gender is appropriate. Of the forty golfers, 20 are men and 20 are women. Divide each group of 20 into groups of 5. Therefore, there are four groups of men and four groups of women.
Men and women are chosen randomly for each group. Brand of golf ball is randomly assigned to a group.
The next step would be to compare the driving distances for each brand of golf ball to answer the question "based on this experiment, which ball is the longest in driving for each gender?"
A matched pairs design is a common form of blocking for comparing just 2 treatments.
In some cases each subject receives both treatments in a random order since order may influence the response. You have probably been the subject of a type of matched pairs experiment when asked by your teachers to take a pre-test....receive instruction...then take a post-test. You are the subject receiving the treatment of education and the comparisons are knowledge before and after instruction.
Or possibly the subjects are matched in pairs as closely as possible, and one subject in each pair receives each treatment. Repeating, in each block, there is either: a SINGLE subject who receives both treatments or a PAIR of subjects, each receiving a different treatment.
Flow charts are useful for describing the order of steps in the design for any type of experiment. These graphs range from fairly simplistic in a completely randomized experiment, shown here, to complex charts for multi-factor experiments perhaps containing blocks that are pictured in your text.
As usual a clearly drawn and labeled picture condenses the meaning down to essential information. For example here is the problem that generated the flow chart pictured.
Suppose a doctor is interested in comparing the anxiety-controlling drug out on the market now (current medicine) with 2 newly available drugs (medicine A and medicine B). A group of 150 patients from a local clinic is available for the experiment. Design an experiment in flow chart form to compare the effect of these three drugs.
IMAGES CREATED BY GAVS