Sunteți pe pagina 1din 7

Correlation

- Correlation

Correlational Studies Correlation Coefficient


Correlation Coefficients — Describe a relationship — Statistic used to
1 between measured describe the relationship
variables between two variables
DESCRIBING ¡ Neither variable is
manipulated — Used when both
variables are on a
RELATIONSHIPS BETWEEN — Association claims
continuous scale
— Correlational findings do
CONTINUOUS VARIABLES —
not imply causation
Variables can be
— Can be used in
correlational research or
continuous or categorical experimental research

Describing a Relationship Four types of relationships between variables


— If both variables are measured continuously
¡ look at relative position of each participant on two
continuous variables
¡ (not comparing two or more groups, the predictor variable is
continuous not categorical)
— Calculate a correlation coefficient -- a number that
quantifies that relationship (r)
¡ r is a number between -1 and 1
¡ Often just referred to as a ‘correlation’

© 2009 The McGraw-Hill C om panies, Inc.

Relationships between variables Correlation Coefficient (r)


6
— Positive Linear Relationship — The correlation coefficient indicates two
¡ Increases in one variable relate to increases in another independent pieces of information
— Negative Linear Relationship — Nature of relationship
¡ Increases in one variable relative to decreases in ¡ Positive Correlation (r>0)
another ¡ Negative Correlation (r<0)

— Curvilinear Relationship ¡ No relationship (r=0)

¡ Increases in one variable relative to both increases and — Strength of relationship


decreases in another ¡ Determined by absolute value of “r”

— No Relationship ¡ Correlation is a number between -1 and +1

1
green blue yellow red purple blue green yellow
green blue yellow red purple blue green yellow
A researcher invited a sample of Person Stroop Calories at Person Stroop Lunch?
college students to a lab at 1:00 p.m. score lunch In another study, the score
to take an attentional control test
called the Stroop, in which color
(seconds to same researcher gave (seconds to
names are presented in different complete) students a Stroop test at complete)
colors. After finishing the Stroop test,
each student wrote down what foods
A 100 550 12:00 p.m. This time, he A 80 Yes
he or she had consumed in the B 120 200 simply asked students B 120 No
previous hour, and the researcher
figured out how many calories each
C 150 210 whether they had eaten C 90 Yes
student had eaten. Make a D 80 600 lunch yet. Make a bar D 110 No
scatterplot of the data in the table
below. What kind of association is
E 85 750 graph of the data in the E 100 Yes
this—positive, negative, or zero? F 90 400 table below. How would F 80 Yes
Does the association seem strong or
weak? Do you notice any outliers?
G 60 575 you describe the G 60 Yes
H 120 325 association here? H 120 No
I 110 350 I 130 No

Sample Data
10

A.Positive relationship (r>0)


B.Negative relationship (r<0)
C.No relationship (r=0)

Sample Data Strength of the relationship


11

— r = 0.60 — r = 0.90

A.Positive relationship (r>0)


B.Negative relationship (r<0)
C.No relationship (r=0)
http://rpsychologist.com/d3/correlation/

2
Strength of the Relationship
(Effect Size)
13 14

— Absolute value of r indicates the strength of the


relationship
Value of r Strength of
Relationship
|r| < .1 None

.1 < |r| < .3 Weak

.3 < |r| < .5 Moderate

.5 < |r| < 1 Strong

15 16

A negative correlation coefficient tells us that A decrease in the number of headaches is


A. if there is a high score on one variable, there is a low associated with a decrease in daily sugar
score on the other variable consumption. What sort of relationship does this
B. if there is a high score on one variable, there is also a describe?
high score on the other variable A. positive correlation
C. the relationship between the two variables is weak
B. negative correlation
D. the relationship between the two variables is strong
C. curvilinear

17 18

The more television a child watches the fewer Moderate amounts of caffeine intake are
social interactions he or she has associated with increased performance, but too
A. positive correlation much caffeine and performance drops off
B. negative correlation A. positive correlation

C. curvilinear B. negative correlation


C. curvilinear

3
Predictor and Outcome?
19 20

Which of the following correlation coefficients — Which is the predictor and which is the
shows the strongest relationship outcome?
A. -0.83 — It depends
B. 1.05 ¡ If there is a time sequence the one that comes first is
the predictor
C. 0.09
¡ There may be theoretical reasoning that indicates
D. 0.26 which is which
¡ It may be a mystery

Interpreting a Correlation Spurious Relationship


21 22
— If two variables are correlated we can conclude:
— X and Y are each related to a third variable and
¡ Perhaps X is related to Y and Y is related to X in a
not directly related to each other
direct relationship
¡ Ice cream sales are positively correlated to drowning
¡ Perhaps both X and Y are related to some other
deaths
variable and are only related to each other in a
¡ Number of storks nesting in Holland is positively
spurious relationship
correlated with human births
¡ Perhaps it is coincidence

— Correlation does not imply causation


¡ It is impossible to tell if changes in X caused changes
in Y (or vice versa), unless the predictor variable was
manipulated (i.e., this was experimental research)

Coincidental Relationship Summary: Correlation Coefficient


23

Strong Mo d e ra te
Weak None Weak Mo d e ra te Strong

-1 -.9 -.8 -.7 -.6 -.5 -.4 -.3 -.2 -.1 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1.0

Negative Positive

— A number between -1 and +1


— Nature of the relationship: positive or negative
— Strength of the relationship: absolute value
How closely the scores cluster around the line
that has the “best fit” with the data.

4
Non-directional (two-tailed) hypotheses Directional (one-tailed) hypotheses:
— There is a relationship between X and Y. — Positive relationship: Higher levels of X are
¡ Null Hypothesis: associated with higher levels of Y
H 0: r = 0 H0 : r = 0 H1 : r > 0
¡ Research Hypothesis:

H 1: r ≠ 0 — Negative relationship: Higher levels of X are


associated with lower levels of Y
H0 : r = 0 H1 : r < 0

28

A child having behavior problems is positively


related to that child being spanked (r=0.46;
r2 =.21). This means that
A. Spanking causes behavior problems
B. Behavior problems cause spanking
C. Both using spanking and having children who
have behavior problems are due to being
aggressive parents
D. None of the above

Evaluating a correlation coefficient


(Statistical Validity)
29 30

A child having behavior problems is positively — Is this a real relationship?


related to that child being spanked (r=0.46; ¡ Statistical significance
r2 =.21). This means that ¡ When do we know that r≠0?

A. As spanking increases we would expect behavior — How big is this relationship?


problems to decrease ¡ Effect size
B. As behavior problems increase we would expect ¡ r2
spanking to decrease
C. As spanking increases we would expect behavior
problems to increase
D. None of the above

5
Is it significant? Statistical significance
31 32

— Is this a real relationship? — A decision rule for rejecting or failing to reject H0


Would you find it again? ¡ We are willing to accept a 5% probability that you
— Statistical significance tells you the probability think you are seeing a real relationship when none
that, if there is no relationship, you would observe exists in the real world. p < 0.05
this pattern by chance. — When there is actually no relationship (i.e.,
— A low p-value tells you that it’s unlikely you would when the NULL is true), you will falsely conclude
have gotten this result if there actually is no that there is an effect 5% of the time.
relationship.

Normal distribution p-value


33 34

r= 0
H0
r= 0
H0

How big is the relationship? Behavior is influenced by many factors


35 36
Value of r Strength of
— Effect Size Relationship — We found a relationship between calories and
— Coefficient of Determination (r2) |r| < .1 None Stroop task performance
Proportion of variance shared by two
— What else might influence your performance?
¡
.1 < |r| < .3 Weak
variables
¡ Child has behavior problems: r=0.46, .3 < |r| < .5 Moderate
r2 =0.21
÷ 21% of the variation in spanking was .5 < |r| < 1 Strong
predicted by the child’s behavior
problems
÷ 21% of the variation in behavior problems
was predicted by spanking

6
37

A child having behavior problems is positively


related to that child being spanked (r=0.46;
r2 =.21). This means that
A. spanking causes 21% of the variation in behavior problems
B. behavior problems cause 21% of the variation in spanking
C. spanking predicts 21% of the variation in behavior problems
D. spanking predicts 46% of the variation in behavior problems

S-ar putea să vă placă și