Managerial Psychology Laboratory, Critical Thinking

Recitation #2 Critical thinking 3 comments • Religious survey • Sexual behavior examples as teaching material Critical thinking • Observing and recording data is the central aspect of social science • • Sometimes, the data at hand is not ideal… • • The question we will start addressing today is what can we learn from different types of data. Correlation & causation • Every day we see in the news claims such as: • • • • • People are happier in better weather Boys who mature later suffer from depression Runners outlive other athletes Kids who breast feed have a higher IQ There is a negative correlation between # of hours watching TV and grades in school • These are reliable statistical relationships, but are they causal? • For each write down the cause for the effect Examples: • Women who exercise regularly have less natural miscarriages. • Can you deduce that exercise help reduce risk of natural miscarriages? Why • Famous conductors in the US live about 7 years longer than the average • Students who come to all the classes do better • Why? What is the story? Correlation & causation • Famous conductors in the US live about 7 years longer than the average. A B A A C B C B A B Interpreting correlations • Sometime we have a very strong causal model • Good weather causes happiness… • Sometime strong models can be wrong • Running prevents heart attacks • Making people run can cause heart attacks... • The “hot hand” So??? • The only way to draw valid conclusions is to have a control group! • In the Uri Geller case we should have tested how many clocks were observed to start working by someone else.. • • In most cases, the only way to have a good control group is to conduct an experiment Self-selection! • Runners select themselves • Famous conductors • Berkeley…. Dep. Men Women Apply Accepted Apply Accepted A 1,000 60% 100 60% B 1,000 30% 1,000 30% Both 2,000 900 1,100 360 = 45% = 33% 5 year cholesterol experiment • P(death|no treatment) = 21% • P(death|treatment) = 21% • What about regularity? • P(death|regularity) = 15% • P(death|non regularity) = 24.6% • But now we have self selection again • For regulars P(D|T)= P(D|no T) ~ 15% • For non regulars P(D|T)= P(D|no T) ~ 26% Experiments • Are not a perfect method • They are also susceptible to some problems • Placebo • The Hawthorne effect • Self fulfilling prophecies • Desire to cooperate Placebo I • In 1794 Gerbi discovered that by squeezing a certain worm between the thumb and the finger of the right hand, and placing the fingers on a painful tooth, the pain went away for a year. • Hundreds of patients were tested and 68% reported no pain for a year! Placebo II • In a study on the effects of birth control, three groups were used • The pill with warning of side effects • Placebo without warning of side effects • Placebo with warning of side effects • Group #2 reported 6% side effects • Groups #1 & #3 reported 20% side effects Placebo III • Old French medical book suggest to use new medicines as fast, when its healing power is the strongest • Probably when the physician believes in it the most Placebo IV • Study on the ability of vitamin C to prevent the common cold • 2 Groups randomly assigned • But it turns out that by opening the capsules the subjects could see if they were in the control or placebo group (the hint was people who dropped out). • When testing only SS who did not know which group they were in, the treatment was not effective. • When testing only SS who did know which group they were in, the treatment was effective Placebo V • The Hawthorne effect • In 1924 a study was done on effectiveness at work • The recommendation was to improve lighting condition • This helped for a while but not long term • Other changes had the same effect …. Placebo -- summary • • • • • Placebo strong effects Threatens the conclusions from many experiments Important to have a good control condition Placebo is real • Opiods • Sensitivity • Attention Self fulfilling prophecies • Teachers who are told that some kids are very smart, find that these kids get better grades • • Researchers who believed that they have genetically inferior rats found them to have a lower performances • • Smart Hans … Desire to cooperate • In many cases subjects want to help the experimenter. • If you tell subjects what is your hypothesis they will help you find it.. • This is why we prefer blind experiment • Even better are double blind experiments • Computer controlled experiments are another approach. Summary • People are bad measurement devices • We observe and infer “rules” where there are non! • • • • Correlations / causation Self-selection Placebo Self fulfilling prophecies • Statistics and research methods are here to protect us against ourselves