STAT-S 300 Lecture Notes - Lecture 9: Institutional Review Board, Dependent And Independent Variables, Church Attendance
Section 4.3 Notes- Using Statistics Wisely 8-28-13
• Scope of Inference
o Can make inferences about pop from results of random samples
o Random assignment helps ensure groups of experimental units are as similar as
possible before treatments imposed
o If one group does much better or worse and difference is too large to be explained
by chance variations, it is statistically significant and must be due to treatments
▪ Can make inferences about cause and effect
▪ Volunteer subjects limit ability to generalize findings to some larger pop
(cannot make inferences about pop)
o Inferences that can be made:
Randomly
assigned?
Yes
No
Randomly
Yes
pop- yes
cause/effect- yes
pop- yes
cause/effect- no
selected?
No
pop- no
cause/effect- yes
pop- no
cause/effect- no
• The Challenges of Establishing Causation
o Well-designed experiment tells that changes in explanatory variable(s) cause
changes in response variable
▪ Tells that something happened in specific experiment in specific
environment to specific individuals- lack of realism can limit ability to
apply conclusions to settings of greatest interest
o Sometimes not practical or ethical to do experiment
▪ Ex. Texting/driving increased risk of accident?
Regular church attendance longer life?
Smoking lung cancer?
▪ Cause-and-effect questions ()- would perform randomized comparative
experiment, but cannot
o Sometimes possible to build strong case for causation in absence of experiment
▪ Criteria (p. 264 for smoking example)
• Association is strong
• Association is consistent
• Larger values of explanatory variable associated with stronger
responses
• Alleged cause precedes effect in time
• Alleged cause is plausible
• Data Ethics (important topic, but not covered on AP Exam)
o “Do not harm”
o Basic Data Ethics
▪ All planned studies must be reviewed in advance by institutional review
board charged with protecting safety and well-being of subjects
▪ All individuals who are subjects in study must give their informed
consent before data are collected
find more resources at oneclass.com
find more resources at oneclass.com