Transcript Chapter 13: Experiments and Observational Studies

Chapter 13: Experiments and Observational
Studies
By: Sahar Abdullah and Jimin Kim
Vocabulary
● Observational study: a study based on data in which no
manipulation of factors has been employed
● Retrospective study: an observational study in which subjects
are selected and then their previous conditions or behaviors are
determined.
● Prospective study: an observational study in which subjects
are followed to observe future outcomes
● Experiment: manipulates factors levels to create treatments,
randomly assigns subjects to treatment levels, and then
compares the responses of the subject groups across treatment
levels
● Random assignment: assigning experimental units to
treatment groups at random.
Vocabulary
● Factor: a variable whose levels are manipulated by the
experimenter
● Response: a variable whose values are compared across
different treatments
● Experimental units: individuals on whom experiment is
performed. Called subjects or participates when the individual is
a human
● Level: the specific values that the experimenter chooses for a
factor
● Treatment: the process, intervention, or other controlled
circumstance applied to randomly assigned experimental units;
the different levels of a single factors or a combination of levels of
two or more factors
Vocabulary
● Principles of experimental design
○ Control
○ Randomize
○ Replicate
○ Block
● Statistically significant: observed difference that is
too large to believe that they are a result of random
fluctuation
● Control group: this group receives the baseline
treatment level; responses provide a basis for comparison
Vocabulary
● Blinding: Keep individuals from knowing how the subjects
are assigned to treatment groups; helps eliminate bias
● Single blind: when those who can influence the results (ex.
subjects) are blind
● Double blind: when those who can influence the results (ex.
subjects) and those who can evaluate the results (ex. treating
physicians) are both blind
● Placebo: a treatment known to have no effect; administered
so that all groups experience the same conditions
● Placebo effect: the tendency of many human subjects to
show a response even when administered a placebo
Vocabulary
● Blocking: isolating the variability attributed to the
differenced
● Matching: reduces unwanted variation
● Designs: in a randomized block design, the randomization
occurs only within blocks
● Confounding: when the levels of one factor are associated
with the levels of another factor so their effects cannot be
separated
Main concept - Principles of Experimental Design
(“Control what you can and randomize the rest.”)
○ Control sources of variation that are not being studied, but will have an
impact on the results. This makes conditions as similar as possible for all
treatment groups. Note: the results of the study cannot be generalized
(i.e., extrapolated) to other levels of the controlled variables.
○ Replicate over as many experimental units as possible. Just as for
simulations, replication reduces random variability in an experiment.
○ Randomize experimental units to treatments. This equalizes the effects
of unknown or uncontrollable sources of variation.
○ Block to reduce the effects of things that cannot be controlled. Blocking
is like stratifying a sample; we separate dissimilar experimental units
into separate blocks, then randomize by treatment within each block. In
effect, the experiment is run separately on each block; the results are
combined across blocks for purposes of study.
Elements of an Experiment
● Plan: State what you want to know.
● Response: Specify the response variable.
● Treatments: Specify the factor levels and treatments.
● Experimental Units: Specify the experimental units.
● Experimental Design: Control, replicate and
randomly assign (also, block if it helps).
● Draw a Diagram
● Perform the Experiment
● Describe the Results: And, determine whether the
results are statistically significant
Homework Problems
35. Some schools teach reading using phonics (the sounds made by
letters) and others using whole languages (word recognition).
Suppose a school district wants to know which method works better.
Suggest a design for an appropriate experiment.
● Treatment: This experiment has 1 factor (reading program), at 2 levels
(phonics and whole language), resulting in 2 treatments.
● Response variable: the reading score on an appropriate reading test after a
year in the program.
● Design: Randomly assign half of the teachers to use phonics and the other half
to use whole language. Then randomly assign students to teachers. There may
be variation in reading score in different schools within the district, as well as
by grade. Blocking by both school and grade will reduce this variation.
Homework Problems
37. A study published in the New England Journal of Medicine (Aug. 2001)
suggests that it’s dangerous to enter a hospital on a weekend. During a 10year period, researchers tracked over 4 million emergency admissions to
hospitals in Ontario, Canada. Their findings revealed that patients
admitted on weekends had a much higher risk of death than those who
went to the emergency room on weekdays.
a) The researchers said the difference in death rates was “statistically
significant.” Explain in this context what this means.
The difference between the death rate on the weekend and the death rate during the
week is greater than would be expected due to natural sampling variation.
b) What kind of study was this. Explain.
It is a prospective observational study. The researchers identifies hospitals in
Ontario, Canada, and tracked admissions to the emergency rooms. This cannot be
an experiment since people cannot be assigned to become injured on a specific day
of the week.
Homework Problems
37. A study published in the New England Journal of Medicine (Aug. 2001) suggests that
it’s dangerous to enter a hospital on a weekend. During a 10-year period, researchers
tracked over 4 million emergency admissions to hospitals in Ontario, Canada. Their
findings revealed that patients admitted on weekends had a much higher risk of death
than those who went to the emergency room on weekdays.
c) If you think you’re quite ill on Saturday, should you wait until
Monday to seek medical help? Explain.
No, it would be foolish to wait since it is likely that confounding variables
accounted for the higher death rate on the weekends. For example, people may
have participated in riskier activities on the weekend.
d) Suggest some possible explanations for this troubling finding.
Drinking may have lead to a higher death rate. Perhaps, more people drank
alcohol, which may have lead to more traffic accidents, and higher rated of
violence.