Chapter 8: Correlation
Which form of study design has higher external validity?
1. Correlational
2. Experimental
Correlational - it is more representative of real-world conditions because you are not changing or manipulating anything about participants' experience.
What form of validity is highest in experiments, and why?
Internal validity is highest because you can control for other variables that might be influential in your results. Experiments allow you to isolate the effects of your treatment/independent variable.
What type of graph best depicts a correlation?
Scatterplot
A rival researcher, Piccolo, decides that he wants to conduct an experiment where he has one group of participants hold their breath for as long as possible before having them complete a questionnaire and the group just completes the questionnaire. What kind of manipulation was conducted?
physiological manipulation (holding their breath)
Goku does a study where he observes that people who eat a special form of bean recover from their injuries faster than those who do not. What type of study is this? He finds a correlation of r = .55, a moderate association. He concludes that the beans must be causing the rapid healing.
What is wrong with Goku's conclusion?
Piccolo decides that he would like to conduct an experiment to see if working out under intense gravity might lead to greater strength gain compared to working out under normal (Earth) gravity. His rival researcher, Goku, found that there was no correlation between exercising in intense gravity and strength gain.
Should Piccolo proceed with his experiment?
He likely should not proceed with his experiment. A precondition for experiments is that there is a correlation between the variables you are interested in studying. If there is not, you are very unlikely to find a relationship through an experiment.
Give an example of when you would use point-biserial correlation.
You use point-biserial correlations when you have 1 continuous variable and 1 categorical dichotomous variable.
An example of when you would use a point-biserial correlation is when you want to examine the association between weight (continuous) and pregnancy status (yes/no; dichotomous).
A participant in an experiment involving watching a TV show and then answering questions did not bring their glasses, so they could not watch the show. The researcher failed to notice this.
What kind of check did the researcher fail to perform?
Goku is back again and this time predicts that dancing will be associated with higher physical strength. He goes out and collects data without manipulating anything. He finds that, yes, dancing does positively predict physical strength (b = .55, p < .001).
What kind of analysis did Goku do?
This was a regression analysis. You can tell because we are interested in prediction, rather than just the association between variables.
Provide an example of a quasi-experimental design.
Your answer should have two parts:
1. The groups are preexisting, you do not manipulate or randomly assign the groups.
2. You administer a treatment to 1 group but not the other.
Ex: testing the effects of a reading intervention (treatment) on classrooms of students (preexisting groups).