[Home]   PSY 101    [Psychology Images] Class 18: Research II: Looking for Links; Evaluating Research-Flaws, Placebo
Last updated: Oct. 5,2021

Looking For Links: Descriptive or Correlational Research

Descriptive or Correlational research can tell us if there is a link between variables, but not about cause and effect relationships

  • Do overweight people exercise less than normal weight people? (weight vs. exercise)
  • Do people in the South live fewer years than people in the North? (location vs. life expectancy)
  • Do angry people have more heart attacks than peaceful people (emotion vs. heart disease)
  • Do step-fathers treat their step-children worse than their natural children (biological relatedness vs. parental care)
  • Do more hours of athletic practice lead to more successful athletic performance? (practice vs. success)

Correlation coefficient (= r): A numerical index between -1 and +1 which expresses the strength of relationship between two variables (correlation coefficient is labeled "r")

[Correlation]
[Negative Correlation] {Zero Correlation] [Positive Correlation]
 Correlation (r) = -1.00   Correlation (r) = +0.05  Correlation (r) = +1.00 
* N = neuroticism, E = extraversion, O = openness to experience, A = agreeableness, C = conscientiousness
As a correlation moves from 0.0 toward +1.0 (more positive), the strength of the relationship increases. Similarly, as a correlation moves from 0.0 toward -1.0 (more negative), the strength of the relationship increases. Correlations near a value of 0.00 indicate that there is little to no relationship between two variables.

Hence, the correlation -0.90 is larger/stronger than the correlation +0.75.   The correlations of +0.63 and -0.63 are exactly the same magnitude but in opposite directions. They are equally "strong". 
Correlations say that there is a relationship, NOT that one variable CAUSES the other. It is possible that both variables are actually caused by a third or fourth variable.

NOT IN BOOK: In doing correlation research, a researcher should have a reason to suspect that a relationship between two variables might exist. In the absence of ANY plausible reason for why two variables might be related, sometimes there are data which show what are called "spurious correlations" ("spurious" means "false, not real, seemingly related but not").

One of the most widely cited examples of a spurious correlation shows almost a perfect relationship between the rate of divorce in the state of Maine between 2000 and 2009 and the per capita consumption of margarine in the U.S. in the same period.
Spurious Correlation

Zombies vs. Political PartiesOr, to use another example, do you think that there might be a relationship between which political party holds the White House and whether Hollywood produces films about Zombies? Well, in one examination of data, it appears that Zombie movies are more likely to be produced when Republicans are President than when Democrats occupy the White House. The correlation of r = ~ 0.30 is marginally significant.

Of course, this statistical relationship does not really indicate any real relationship.

There is a large list of more and often very funny spurious correlations at this website.

Other forms of Descriptive Research


Flaws: Evaluating Research

Sampling Bias: Is the sample representative of the population under review?

Bias can be introduced by
Consider the infamous 1936 Literary Digest "poll" which predicted that Alf Landon would beat Franklin D. Roosevelt in the presidential election by 57% to 43% (and win a landslide 370 electoral votes!) The magazine had contacted 10 million Americans (whose names were drawn from long lists of people who owned cars or telephones). The magazine got back over 2 million responses. Of course, in the actual election, Roosevelt beat Landon by 61% of the votes (523 electoral votes vs. 8 electoral votes for Landon)
What did the Literary Digest do wrong?
See The First Measured Century (PBS): George Gallup and the Scientific Opinion Poll

Placebo Effects: Changes in a person's behavior which come from the EXPECTATION of change, rather than the ingredients or components of the treatment they receive.

Distortions in Self-Report Data: Bias introduced by participants who respond in ways that do not reflect their actual behavior, beliefs, judgments, etc.

Experimenter Bias: Bias introduced when a researcher's expectations or preferences influence the outcome of the research. This may be done without the research ever realizing that he/she is affecting the outcome.

============================

Looking at Ethics

Psychology studies human beings and we humans are a very complex reality to study. We are often smart enough to understand the purpose of a research study in ways that would make that study invalid. Experimenters try to get people to respond honestly, but participants might distort their responses for any number of reasons.

In the last class, I gave you an example of researching whether faculty members might have an unconscious bias against women. Notice that the participants were not told that they were being studied for bias before their responses were collected. Indeed, if you asked the participating faculty directly, they would almost certainly have denied that they had any bias whatsoever. So, the experiment disguised how bias was going to be measured by NOT telling the participants what the independent variable was, that is, the applicant was either male or female. Is such deception justifiable or ethical?

The question of deception

Animal Research

Ethical Principles in Research with Human Subjects

 
Reference

Downs, A.C., & Lyons, P.M. (1991). Natural observations of the links between attractiveness and initial legal judgments. Personality and Social Psychology Bulletin, 17, 541-547.

Gunnell, J. J., & Ceci, S. J. (2010). When emotionality trumps reason: A study of individual processing style and juror bias. Behavioral Sciences and the Law, 28, 850-877. doi: 10.1002/bsl.939


[Home]
This page originally posted on 1/28/04 and updated on Oct. 5, 2020