PCE- Designing Experiments Lesson

Designing Experiments Lesson

If we want to observe individuals and record data without intervention, we conduct an OBSERVATIONAL STUDY. However,   if we want to examine a cause and  effect  relationship, we conduct an EXPERIMENT.     The individuals on which the experiment is done are called "experimental units."   If the units are people they are called SUBJECTS.  The experimental treatment we apply to the units is called the TREATMENT.  

The explanatory variables (causing a change in the other variables) are called FACTORS.    The factors may be applied in different LEVELS.   Here is an example to see how to identify all the parts in a real experiment.

Is diet or exercise effective in combating insomnia?

Some believe that cutting out desserts can help alleviate the problem while others recommend exercise.
Forty volunteers suffering from insomnia agreed to participate in a month-long experiment.
Half were randomly assigned to a special no-desserts diet; the others continued desserts as usual.
Half of the people in each of these groups were randomly assigned to an exercise program, while the others did not exercise.   Those who ate no desserts and engaged in exercise showed the most improvement.

subjects:   people suffering from insomnia
factors:    TWO factors   - desserts and exercise
levels:  not present in this question but would be things like varying amounts of exercise and restricted deserts
treatments:    FOUR treatments - dessert and exercise, no dessert and exercise, dessert and no exercise, no dessert and no exercise.
response variable:    less time to fall asleep or more hours of sleep

A treatment is defined as the experimental CONDITION.   Treatments are the levels of a factor.

 

Example of a one factor experiment:

Suppose a gardener wants to find out the amount of water needed so that her flowers will look the best. The amount of water that is given to each group of flowers is the treatment.   There may be as many as five levels of watering:   none, light, medium, heavy, and very heavy. Treatments in a multifactor experiment are a COMBINATION of levels of the factors.

Example

Suppose a gardener wants to find out the amount of water needed so that her flowers will look the best. She is ALSO concerned about the best time to water the flowers. The time and the amount of water that is given to each group of flowers are the treatments.   Imagine our first example broken down into a.m. and p.m. producing a total of 10 different treatment options.

BEWARE the lurking variables!   When designing an experiment we want to MINIMIZE the effect of lurking variables so that our results are not biased.   Lurking variables are possible confounding issues that are not specifically part of the experiment. Because we may not be able to identify and eliminate ALL lurking variables, it is essential that we use a CONTROL GROUP.   Using a control group is an attempt to minimize or remove as many lurking variables as possible.   The control group gets a "fake" treatment to counter the PLACEBO EFFECT where subjects have a positive response simply by thinking they are getting the treatment.  The HAWTHORNE EFFECT is another common response where people knowing they are part of an experiment change their normal behavior or act in different ways.   Having a control group allows us to compare the results of the treatments with "confidence."   For example, pharmaceutical manufacturers spend lots of money on research and development (R and D) to test the efficacy of a new drug versus the current drug. Because of the large financial expenditure it is important that the results of these experiments be reliable and control groups play an important role in that.

A blind experiment is designed so that individuals do NOT KNOW whether they are the so-called "test subjects" OR members of an experimental control group. Double-blind describes an especially stringent way of conducting an experiment, usually on living, conscious, human subjects, i.e. those with reasoning ability.   In a double-blind experiment, NEITHER the individuals NOR the researchers know who belongs to the control group.   Only after all the data are recorded (and in some cases, analyzed) may researchers be permitted to learn which individuals are which.

 

Performing an experiment in double-blind fashion is a way to lessen the influence of prejudices and unintentional physical cues on the results.  

How might we define a triple-blind experiment?

If it even existed, triple-blind would describe a way of conducting an experiment where NONE of the parties are aware what is going on...the individual subjects, the researchers administering the treatments and recording the data, and those that originally designed the experiment and will be forming the conclusion.   There is no such thing as triple-blind since it would be impossible for the researcher to not have knowledge about his own experiment.   A little statistics humor!

Putting this all together, here are the steps involved in designing an experiment.

Step 1: 

Choose the treatment(s)
Identify the factors and levels
Assign a control group

Step 2:     

Assign the experimental units to the treatments
Block units if applicable (place similar units in each treatment group)
Randomize (randomly assign units to each treatment group)

In the next lesson you will learn that there are specific criteria necessary for an experiment to be WELL-DESIGNED.

 

IMAGES CREATED BY GAVS