PHIL2420 Lecture Notes - Lecture 3: Photorealism, Citric Acid Cycle, Publication Bias
1 Jun 2018
School
Department
Course
Professor
PHIL 2420
Critical Thinking
March 15, 2018
WEEK 3
Causality (Revisited), Models & Analogies
Mill’s methods don’t give mechanical way to test hypotheses
Method though trial and error and critique
Smoking and lung cancer
• Correlation vs Causality
• Has there ever been a controlled trial?
o A controlled trial would be unethical
o A controlled trail would take a long time
▪ 50+ years for effects to show up
Establishing Causality without RCTs
• Observational Studies (Epidemiology)
o Retrospective (case-control)
o Prospective (cohort)
*Evidence Based Medicine
• Asserts that RCTs are the Gold Standard
• We like to think of our medicine based on science
o Systematic Review
▪ Evaluate all evidence on a particular question
▪ Hierarchy of Evidence
• Case Study
•
o Bunch of RCTs is the best standard for evaluating evidence
• It’s not the gold standard. Not all results based on a RCT are valid.
• Some observational studies are better than some RCTs.
• IT DEPENDS on the overall design QUALITY of the study in question
• Grossman & Mackenzie, 2005
Retrospective Study
• Random sample of human beings
• Two groups
o 1 – with suspected effect
o 2 – with no suspected effect
• See if there’s a difference in rate of exposure to the suspected cause between the
two groups
find more resources at oneclass.com
find more resources at oneclass.com
Prospective Study
• Random sample of human beings
• Two groups
o 1 – exposed to the suspected cause
o 2 – not exposed
• See if there’s a difference in rate of suspected effect over time between the two
groups
Confounding Variables
• Extraneous variable – potential cause that was not controlled in a study
• Independent variable – suspected cause
• Dependent variable
• Confounding variable – causal and correlated with IV
o Causality implies correlation
o Correlation does not imply causality
Observational Studies
• Advantages
o Useful when an RCT would be unethical
o Can utilize existing data
o Easy to attain large sample sizes
• Disadvantage
o No control for confounding variables
Prospective VS Retrospective
• Retrospective
o Designs can be conducted quickly
o Prone to unsuitable available data and/or recall bias
▪ How many cigs did you smoke at 17 years old? To a 78-year old man
▪ May not have data you’re really looking for
• Prospective
o Allow for some choice in variables
o Can require long-term follow up
o Harder to do
Horn-Hammond Study – NOT ENOUGH TO PROVE CAUSALITY
• Cohort
o 188,000 American men
o 50-69
o 10 states
• Study period – 1952-1955
• IV – number of cigs smoked
• DV – number of cigs smoked
• Method – questionnaire, follow up, death certificates
• Result
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
Mill"s methods don"t give mechanical way to test hypotheses. Smoking and lung cancer: correlation vs causality, has there ever been a controlled trial, a controlled trial would be unethical, a controlled trail would take a long time, 50+ years for effects to show up. Establishing causality without rcts: observational studies (epidemiology, retrospective (case-control, prospective (cohort) Not all results based on a rct are valid: some observational studies are better than some rcts, grossman & mackenzie, 2005. It depends on the overall design quality of the study in question. Retrospective study: random sample of human beings, two groups, 1 with suspected effect, 2 with no suspected effect, see if there"s a difference in rate of exposure to the suspected cause between the two groups. Prospective study: random sample of human beings, two groups, 1 exposed to the suspected cause, 2 not exposed groups. Confounding variables: see if there"s a difference in rate of suspected effect over time between the two.
