Understanding Experimental Methods September 13, 2017
Different Kinds of Experiments Laboratory Experiments Field Experiment Natural Experiment
Confounding Variables Demand Characteristics -Hawthorne effect -Single blind control Researcher Bias -Double blind control Participant Variability Artificiality
Correlational Studies -Positive correlation -Negative correlation -Bidirectional ambiguity -Correlation and causation
Correlations in the Headlines
Problem 1
There are three independent variables There are three independent variables. We cannot know from this whether the diet has a more direct effect than the other two variables - or whether it has no effect whatsoever. It may also be that the three variables must all be present and need to interact in order to produce the desired behaviour. The word "prevention" is problematic as no cause and effect can be established. The teens had "criminal records." That does not necessarily mean "violence." What type of fish was eaten? And for how long? Was it fish high in Omega-3 fats? How did the researchers control whether or not the other group exercised? How is exercise defined? Cognitive stimulation? How were they measured? There is no control (that we know of) of the family history or environmental factors. We do not know if the three-year-olds were randomly allocated to conditions.
Problem 2
Problem 2 There is an assumption here that people who drink beer (versus other alcohol) make poor choices, have unprotected sex and get gonorrhea. That may be a faulty assumption. There is no evidence that there was a higher rate of gonorrhea among beer drinkers prior to the study. There may have been a public health campaign some time between 2001 and 2007 that influenced the sexual behaviour of the teens. There is no comparison made to states where the price of beer did not increase. There is no indication whether this is a steady decline or the average decline over the seven years. This is correlational research, so no cause and effect can be established. It is assumed by the stats that all people who get gonorrhea drink beer. The cost increase is minimal. Do the stats reflect a decrease only in the lowest classes? If not, this may not be a logical conclusion.
Problem 3
Problem 3 What is the definition of "lucky?" Some people would say that if they have good health and a good job, they are "lucky." Others may only see "luck" as winning the lottery. The definition can also be culturally relative. This was an online survey. So, the results may not be representative of the general population. It may, however, be more cross-cultural. We don't know the number of people that responded for each of the months. If only 2 people, for example, had November birthdays, then 1 person saying that they were lucky meant that only 50% felt lucky. There are 40,000 respondents, so that is a good number, but the data may still be highly skewed. The difference in the percentages is rather small and may not be statistically significant. Could the month in which this was asked have affected the results?
Problem 4
Problem 4 First and foremost, not every religion believes in hell. Jews, for example, do not have hell as part of their religion. Neither do Buddhists. Or Atheists. There is a question of how the researchers determined whether people "believe in hell." Did they simply count the number of Christians in the US? Did they do an online survey and if so, who answered it? How many people were actually surveyed in order to extrapolate to 71% of the population? There is the problem of generalizing from smaller samples to target populations with surveys in diverse populations like the United States. The study is correlational and cannot establish causation. There is no discussion about the resources available to the country. This could be another important variable. Only 35 countries were reviewed. There are currently 192 countries represented in the United Nations. Were these countries randomly chosen?
For tomorrow... Please bring your Unit One review questions to class