I. Correlational Research
Based on Observations (Empirical)
-
Shows how One thing (A) relates to another
(B) across Individuals.
-
e.g. Shoe Size relates to Height
-
e.g. High School GPA relates to College
GPA
-
e.g. Class Attendance relates to Grades
on Tests
-
e.g. Mother's Cocaine Use relates to Child's
Health
II. Observe A and B for a number of Individuals
-
Obtain Quantitative Measures of A and
B for each.
-
Example from the Amazon:
-
People living along Rio
Tapajós
-
Exposed to Mercury in fish from the river.
-
Scatterplot:
X-axis is Level of Mercury.
-
Scatterplot:
Y-axis is measure of performance
-
Each individual is represented as a point.
-
Is there a Relation? Or is there no Relation?
-
Is it a Positive Relation? Or Negative?
-
How STRONG is the relation?
-
Correlation Coefficient (CC)
(No=0.0, Strong=1.0 or -1.0)
-
Coefficient of Determination (%
of variation)
Square of CC: 1.0*1.0 =1.0 (All)
0.67*0.67 =0.45
0.5*0.5 =0.25 (1/4)
III. Correlation Provides a Measure
of Predictability
-
If you know A, how well can you predict
B?
-
Does that mean A causes B??? No.
-
A->B? or B->A? or C causes both A and
B?
-
e.g. Foot size -> Number of Words in Children?
-
No. Age increases Both.
IV. vs Experimental Research.
A. Problems for Correlational
Research.
-
Can't tell Cause/Effect like Experiments
can.
-
Can't easily eliminate CONFOUNDING
VARIABLES (two variables are confounded if you cannot distinguish their
effects).
Example of the problem: Mother's cocaine
use correlates with poor health of child. Due to cocaine use? Poverty?
Social isolation? Low education? Chaotic living condition? Ö All these
"other factors" are confounded with cocaine use and therefore we cannot
"blame" cocaine as the cause.
B. Benefits for Correlational
Research.
-
A quantitative (numerical) measure
of the relation.
