Chapter 5 Correlation & Causation

Chapter 5: Correlation & Causation A large-scale study in Taiwan of factors related to contraceptive use revealed that the best predictor of birth control use was how many electrical appliances were in the home. Obviously, correlation does not imply causation, but sometimes the link is not as obvious as this. The case of Karl Pearson: Pearson was a British statistician who developed Pearson’s r at the turn of the 20th century. He calculated the correlations between many biological variables and was a strong believer in heredity & eugenics. He found that the link between parents & child tuberculosis was .50. He thought that virtually every imaginable human characteristic was primarily a product of genetic influence and used this belief to justify his prejudices. He said that Jews living in slums were “innately dirty.” As for the link between parent-child TB, he saw the .50 correlation as being proof of heredity and dismissed sanitation campaigns against the disease on the grounds that heredity made such attempts futile. His preexisting bias ended up hurting people because improved sanitary conditions greatly reduces the rate of TB. The case of pellagra: Pellagra was a disease affecting thousands of Americans in the South in the early 1900s. It was characterized by dizziness, lethargy, running sores, vomiting, and severe diarrhea. The disease was thought to be caused by a living microorganism of “unknown origin,” and evidence linked it to unsanitary living conditions. Scientists found that homes in Spartanburg, SC, that had indoor plumbing and good sewerage did not have pellagra. A doctor who doubted this was Joseph Goldberger, who, at the Surgeon General’s direction, conducted several investigations of pellagra and decided that it was caused by inadequate diet of too many carbs and not enough protein and dairy. (Too much corn, grits, and mush, and not enough milk, eggs, and meat). Goldberger thought the correlation arose because families with sanitary plumbing were better off and able to eat better foods. Goldberger’s thought was still just a correlation, though, not a causal factor. Why do we trust his judgment and not the other possibility about unsanitary conditions causing the disease? Because he ate the excrement of pellagra victims! In other words, Goldberger actually manipulated the critical variables, turning it into an experiment rather than a correlational study. Goldberger was so confident that pellagra was not contagious and not transmitted by the bodily fluids of the victims that he did the following… 1) Injected himself with the blood of an infected patient 2) Inserted secretions from the throat and nose of a victim into his own mouth, and 3) “Finally, he selected two patients—one with scaling sores and the other with diarrhea. He scraped the scales from the sores, mixed the scales with four cubic centimeters of urine from the same patients, added an equal amount of liquid feces, and rolled the mixtures into little dough balls by the addition of four pinches of flour. The pills were taken voluntarily by him, by his assistants, and by his wife.” (Bronfenbrenner & Mahoney, 1975, p. 11). 4) Then he decided to test his own causal mechanism by selecting two groups of prisoners from a Mississippi state prison (who supposedly volunteered). One group was randomly assigned to a balanced diet, and the other was assigned to a high-carb, low-protein diet that he suspected was the cause of pellagra. Within 5 months, the lowprotein group was ravaged by pellagra while the other group showed no sign of the disease. Goldberger had a tough fight politically against those who wanted to deny the existence of poverty and against those who wanted to advance hereditarian explanations for illnesses, but his hypothesis was eventually accepted because there was no denying his findings. The pellagra case represents the dangers of mistaken inferences from correlational studies. Sometimes, correlational evidence is all we can have. And if you’re just using it for the purpose of prediction (instead of finding the cause), it’s really all you need. 6 The Bidirectionality Problem: Eye Movements and Reading: Researchers figured out 100 years ago that there was a correlation between eye movement patterns and reading ability. Poorer readers make more erratic movements, display more regressions (movements from right to left), and make more fixations (stops) per line of text. On the basis of this correlation, some educators thought that deficient oculomotor skills were the cause of reading problems, so expensive “eye movement training” programs were administered to elem. School children. We now know that the causal relationship runs in exactly the opposite direction. Erratic eye movements don’t cause reading problems…it’s the other way around. Slow recognition of words and difficulties with comprehension leads to erratic eye movements. When kids are taught to recognize words efficiently and comprehend them, their eye movements become smoother. Training kids’ eye movements do not help at all. The third-variable problem: the fact that the correlation between two variables may not indicate a causal path in either direction but may arise because both variables are related to a third variable that has not been measured. Spurious correlations: correlations that arise NOT because a causal link exists between the two variables that are measured but because both variables are related to a third variable. Example: Public vs. private schools: During the 1980s, debates raged about which was better…private vs. public schools. At the crux of the private school argument is the fact that student achievement in private schools exceeds that in public schools. This fact is empirical and not in dispute. The problem is the conclusion that the education received in the private schools causes the superior test scores. We need a more sophisticated statistic to evaluate the true relationship because we know that educational achievement is related to many different indicators of family background (education of the parents, parental occupation, SES, the # of books in a home, etc.) that are also related to the likelihood of sending a child to private school. We need to be able to rule out family environment as a 3rd variable, and we can use multiple regression and path analysis to do that. After conducting these analyses and controlling for these factors, several researchers have found no relationship between private vs. public education. The reasons why private school students perform better are due to family environment and different preexisting cognitive abilities in private school students. **If the original correlation remains even after other alternative explanations have been factored out statistically, it tells us that the original correlation was not due to a spurious relationship with those particular third variables. It does not remove the possibility of a spurious relationship due to some other 3rd variable that wasn’t tested. Example #1: prediction of who goes to college: Thomas, Alexander, and Eckland (1979): found that the probability that a high school student will attend college is related to the SES of the student’s family. This suggests that opportunities for success in life are determined by a person’s economic class. Before we jump to this conclusion, though, we need to examine other correlations for spuriousness. One potential spurious variable is academic ability. This was factored out, and the original correlation remained, telling us that the academic ability was not a factor. Example #2: High schoolers who have a job: Developmental psychologists have found that more work experience is related to poorer school achievement and higher delinquency. Alternative explanations are that teens who choose to work may have different home backgrounds or lower academic ability to begin with. However, researchers have factored these out, and the negative correlation remains. Selection bias: refers to the relationships between certain subject and environmental variables that may arise when people with different biological, behavioral, and psychological characteristics select different types of environments. Selection bias creates a spurious correlation between environmental characteristics and behavioral-biological characteristics. 7 Example 1: Quickly name the state with the highest incidence of deaths due to respiratory illness. The answer is Arizona…because people with respiratory illnesses move there in hopes of cleaner air. Example 2: Correlational research shows that SAT scores were higher in Mississippi than they were in California (at least during the 1980s). How? Because 42% of California students take the SAT, whereas only 3% (the cream of the crop) take them in Miss. Similarly, a statistic shows that the more money poured into education, the LOWER the average SAT scores. That’s because good educational systems tend to produce more students who want to go to college, resulting in more taking the SAT. There’s more variance then. 8