chance is vital in statistical design.
Random sampling and randomized comparative experiments are extremely important and effective statistical practices.
observational study vs experiment
Observationalstudy - Observe individuals
- Measure an area of interest - Do no influence response
� Experiment
- Treatment imposed on individuals in order to observe
responses
- Evidence for causation
observational study
� A group of 60 exercisers are classified as walkers and runners. A study over time is done to determine who is healthier.
A group of 60 volunteers who do not exercise are randomly assigned to a fitness program. One group of 30 is enrolled in a walking program, the other group of 30 is put into a running program. After a period of time, the 2 groups are compared for fitness.
- Experiment because the researcher imposed a treatment and measured the effect.
control
overall effort to minimize variability in the way experimental units are obtained and treated
experimental units
Individuals on which experiment is done (people to subjects)
treatment
A specific experimental condition applied to experimental units.
factor
Explanatory variable -treatment
placebo
A dummy treatment that can have no physical effect.
Control group
receive dummy treatment. Helps experimenter control effects of lurking & confounding variables.
Completely randomized experimental design
All experimental units are allocated randomly among the treatments.
Statistically significant observation
an observed result too unusual to be an outcome determined by pure chance.
Three principles of experimental design
control, randomization, replication
control
Needed to counter effects of lurking & confounding variables
- Simplest form is comparison (2 or more treatments)
- Types include control group, placebo, blinding
replication
Do experiment on many experimental units to reduce chance variation in
the results.
Randomization
- Subjects assigned to treatments by chance
- Creates groups that are similar, except for chance variation.
Types of Experimental Design
� Completely Randomized Design
� Block Design
� Matched Pair Design
Completely Randomized Design
Three essential elements
- random allocation of subjects to treatment and
control groups
- administration of different treatments
- comparison of the outcomes between the groups
Block Design
� A block is a group of experimental units or subjects that are known before the experiment to be similar in some way that affects the response to treatment (age, gender, race, health)
� Perform a completely randomized experiment within each block
� Block
Block Design Mantra
� Control what you can
� Block what you can't control � Randomize the rest
Matched Pair Design
� A special case of a randomized block design.
� Used when the experiment has two treatments subjects can be matched in pairs that are similar, based on some blocking variable.
� Within each pair,subjects are randomly assigned to different treatments.
Matched Pair Design
� Aspecialcaseofarandomizedblockdesign.
� Usedwhentheexperimenthastwotreatments subjects can be matched in pairs that are similar, based on some blocking variable.
� Withineachpair,subjectsarerandomlyassignedto different treatments.
Variation of Matched Pair Design
� One subject and impose both treatments on the same subject
� Randomlychoosewhichtreatmentdotofirst
� Measure the difference between the treatments
� Addedlevelofcontrolbecausethedifference between treatments is measured (improvement)
Placebo Effect
� Phenomenonwherepatientsgetbetterbecausethey expect the treatment to work event though they have been given a placebo (fake treatement).
Control: Blinding
Blinding
- Subject does not know which group they are in
� Double Blind Experiments
- Neither the subject nor the researcher knows
which group they are.
� Help control the placebo effect
Cautions
Need to treat all the experimental units identically in every way except for the actual treatment
� Requires attention to detail
� If using blocking be sure to randomize treatments within each block