CORRELATIONAL RESEARCH AND SMALL N DESIGNS

Outlier - score that is radically different from other data points

Outliers can have a big impact on correlations. It's always important to look at a scatterplot of your data to see if you have outliers.

 

Terms used in correlational research:

predictor variable

analogous to Indep Var (but you=re not manipulating it)

used to predict outcome on criterion

criterion variable

analogous to Depend Var (but you don=t really know if it is caused by Indep Var)

variable predicted by predictor variable

Ex: SAT scores are a predictor variable; College GPA is a criterion variable (correlational study).

 

correlation - the extent to which two continuous variables covary

correlations range from -1.0 (perfect inverse relationship) to 1.0 (perfect direct relationship). The predictive value of a correlation depends on its distance from zero, regardless of whether it's positive or negative.

Correlation does not equal causality!

Variable A could cause Variable B

Variable B could cause Variable A

or

Some third variable (C) could cause both Variables A and B.

 

r (symbol used for correlation)

 

curvilinear relationship - when the relationship between two variables is not well represented by a straight line.

Classic example: stress and performance. At low stress levels, increasing stress improves performance, but at high stress levels, greater stress tends to produce worse performance. If we were to graph this relationship, with stress on the X axis and performance on the Y axis, the shape would be an inverted U, with the highest performance in the middle of the U.

The correlation coefficient for such a relationship would likely be close to 0.

 

 

small N designs:

rare cases

clinical interventions

develop hypotheses

idiographic approach

behaviorism studies

idiographic approach: centered within one individual; emphasizes a coherent explanation of one person.

nomothetic approach: seeks to explain across individuals, emphasizes an explanation that generally applies to many people.

 

problems with small N

one-shot

bias

lack of generalization

 

Examples of Small N Designs:

ABAB (or withdrawal)

Multiple baseline

 

applied research: research with the goal of trying to solve some immediate real life problem

basic research: research with the goal of describing, predicting, and explaining fundamental principles of behavior

 

Examples of quasi-experimental designs:

Interrupted time series -

Ex: Does the number of drunk driving accidents go down after the drinking age has been raised? We would examine the number of accidents in the years before and after the age change.

Non-equivalent control -

Ex: Is the number of drunk driving accidents less in the state that raised its drinking age than in another state that didn't raise its drinking age?