Вы находитесь на странице: 1из 42

Week 9

From Correlations to
Multiple Differences

Today
Correlations
Differences between more than 2 means
Part 1

Correlation
Correlation Coefficient (r):

between -1 and 1
0 indicates no correlation at all
-1 means perfect negative correlation
1 means perfect positive correlation

Coefficient of Determination (R2):


between 0 and 1
is the % of explained variation
R2 = 0:
you cannot explain, or predict, the value of your dependent
variable if you know the value for your independent variable

R2 = 1:
you can perfectly explain, or predict the value of your
dependent variable from your independent variable

Can we predict the score for the


dependent variable, for an
individual with a score of 5 for the
independent variable?

10

Dependent Variable

9
8
7
6
5
4
3
2

R2=1.00

1
0
0

Independent Variable

10

Can we predict the score for the


dependent variable, for an
individual with a score of 5 for the
independent variable?

10

Dependent Variable

9
8
7
6
5
4
3
2

R2=0.90

1
0
0

Independent Variable

10

Can we predict the score for the


dependent variable, for an
individual with a score of 5 for the
independent variable?

10

Dependent Variable

9
8
7
6
5
4
3
2

R2=0.80

1
0
0

Independent Variable

10

Can we predict the score for the


dependent variable, for an
individual with a score of 5 for the
independent variable?

10

Dependent Variable

9
8
7
6
5
4
3
2

R2=0.60

1
0
0

Independent Variable

10

Size of Correlation
(Cohen, 1988)

WEAK (0.1 - 0.3)

MODERATE (0.3 - 0.5)

STRONG (> 0.5)

Correlation
Is the correlation significant?
Almost ALL correlations will be non-zero
Could we find a similar or larger correlation just by
chance?
Is r large enough for it to be unlikely that the data are
really unrelated?
SPSS calculates that probability
If p < 0.05 we reject the null-hypothesis and conclude
there is a SIGNIFICANT correlation

Cautions
Be careful with:
non-linear relations

Cautions
Be careful with:
non-linear relations
heterogeneous samples

Cautions
Be careful with:
non-linear relations
heterogenous samples
sub-samples

Cautions
Be careful with:
non-linear relations
heterogenous samples
sub-samples

Cautions
Be careful with:

non-linear relations
heterogenous samples
sub-samples
outliers

Cautions
Be careful with:

non-linear relations
heterogenous samples
sub-samples
outliers

r = 0.90

Cautions
Be careful with:

non-linear relations
heterogenous samples
sub-samples
outliers

r = 0.09

Talking about correlations


In a results section, mention:
whether the correlation was significant
if not, thats all you have to say
no need to state the r-value
no need to state (p>0.05)

if so: was it positive or negative?


in brackets: (r-value; level of significance)
if p>0.05:
No significant correlation was observed between variable A and variable B.

if p<0.05:
There was a significant negative correlation between variable A and variable
B (r=-0.56; p<0.05).

Probably (p< .05?) enough


about correlations...
Lets go back to testing for
differences

35

40

30

35

performance (arbitrary units)

Performance (arbitrary units)

Correlation or Difference?

25
20
15
10
5

30
25
20
15
10
5
0

0
1

Confidence (scale from 1-10)

10

Confidence (arbitrary units)

10

11

Dependent variable

Dependent variable

Correlation or Difference?

3
2
1

3
2
1
0

0
0

2
3
4
Independent variable

2
3
4
Independent variable

Multiple Differences?
If you have (e.g.) 3, 4, 5 or more conditions of an
independent variable:
For example, different exam conditions
you will have to use a statistical test which compares
>2 conditions at the same time
is there a significant difference between ANY of the conditions?
NOTE: For conditions also read groups or samples

Testing for differences


For comparing > 2 INDEPENDENT samples:
One-way ANOVA (sometimes called unrelated)

For comparing > 2 PAIRED samples:


Repeated measures ANOVA (sometimes related)

ANOVA
ANalysis Of VAriance
To determine whether one or more of the
samples (e.g., conditions) is/are different
from the others compare the:
between-group VARIANCE and the
within-group VARIANCE

within-group variance

between-group variance

within-group variance

between-group variance

within-group
variance

between-group variance

within-group
variance

ANOVA: outcome
If ANOVA results in p<0.05
this means there is a difference between the groups
but between which ones???

An additional test is required to identify


WHICH samples are different from each
other
Post hoc test:
Bonferroni comparisons
SPSS will only do this for you if you tell it to!

Procedure:
Test:
One-way unrelated ANOVA
p=0.14
No significant differences between any of the samples
14

Score (units)

12
10
8
6
4
2
0
<18

18-39

39-65

Age-group (y)

>65

Procedure:
Test:
One-way unrelated ANOVA
p<0.001
Post hoc Bonferroni comparisons:
14

Score (units)

12
10

a
c

8
6

a, b, c

4
2
0
<18

18-39

39-65

Age-group (y)

>65

Procedure:
In some situations, you can still use asterisk system
e.g., if middle variable differs from other two variables
Add appropriate text beneath table
*** Variable Y significantly different from Variables X and Z (p < .001)
35
30

Pre-exercise
5 minutes post-exercise

**

30 minutes post-exercise

Score

25

***

20
15
10
5
0
Stress

Arousal

Hedonic tone

SPSS: repeated measures ANOVA

Looking for:

a difference
using a parametric test
for paired samples (= repeated measures)
between more than 2 means

Look up the p-value in the SPSS output


Not easy to find, so note down where to look!
if p<0.05:
there is a difference between at least 2 of the means, and maybe
more

look up pair-wise comparisons to find out WHICH


means are different from eachother

Pair-wise comparisons
Sample 1

1 vs. 2
1 vs. 3
2 vs. 3
SPSS:

Sample 2

Sample 3

Reminder - Statistical
analyses section
Descriptive statistics
Inferential statistics - What tests were used?
for what purpose
to test for differences or a correlation?

between what samples


for which variables

Example
Statistical analyses
Pearsons correlation coefficient was used to determine the correlation
between maximal oxygen uptake capacity and 10-mile cycling timetrial performance. Mann-Whitney U-tests were used to examine the
effect of gender on autistic quotient and emotional intelligence.
Significance was accepted at p<0.05.

Results section
There was a significant negative correlation between maximal oxygen
uptake capacity (VO2max) and time taken to complete a 10-mile
cycling time trial (r=-0.50; p<0.05; Figure 1).
There were no significant gender differences for autistic quotient
(Figure 2A) or emotional intelligence (Figure 2B).

Spot the difference:


the difference between males and females for autistic quotient
and emotional intelligence
the difference between autistic quotient and emotional
intelligence for males and females

English
There is a difference between money and the day of the
week
There is a difference between the amount of money you
have at the beginning and the end of the week
Only one of the above makes sense...and allows you to
report the answer to the (probable) question
You have more money at the beginning of the week than
at the end of the week

Q1.

Gender differences in IQ
Group Statistics
Gen

Mean

Std. Deviation

Std. Error Mean

1.00

24

116.25

12.519

2.555

2.00

15

115.00

12.739

3.289

IQ

Independent Samples Test


Levene's Test for Equality of Variances
F

Sig.

t-test for Equality of Means


t

df

Sig. (2-tailed)

Mean Difference

Std. Error

95% Confidence Interval of the

Difference

Difference
Lower

Equal variances assumed

.882

.354

Upper

.301

37

.765

1.250

4.148

-7.155

9.655

.300

29.466

.766

1.250

4.165

-7.263

9.763

IQ
Equal variances not assumed

Results indicated that there were no significant gender differences in IQ in females


(M = 116.25, SE = 2.56) and males (M = 115.00, SE = 3.29), t(37) = .30, p = .77,
two-tailed).

Q2. Differences in shoe size between psychology students and sport and
exercise science students

Group Statistics
Coursecode

Mean

Std. Deviation

Std. Error Mean

1.00

23

6.52

2.150

.448

2.00

16

7.81

2.738

.684

ShoeSize

Independent Samples Test


Levene's Test for Equality of Variances
F

Sig.

t-test for Equality of Means


t

df

Sig. (2-tailed)

Mean Difference

Std. Error

95% Confidence Interval of the

Difference

Difference
Lower

Equal variances assumed

1.226

.275

Upper

-1.648

37

.108

-1.291

.783

-2.878

.296

-1.577

27.220

.126

-1.291

.818

-2.969

.388

ShoeSize
Equal variances not assumed

Results indicated that the difference in shoe size between Psychology students (M
= 6.52, SE = .45) and Sports and exercise students (M = 7.81, SE = .68) was not
significantly different, t(37) = .30, p = .11, two-tailed.).

3. Differences in the time taken to complete the first and second marathons

Paired Samples Statistics


Mean

Std. Deviation

Std. Error Mean

Marathon1

226.82

39

61.204

9.800

Marathon2

204.69

39

52.285

8.372

Pair 1

Paired Samples Test


Paired Differences
Mean

Std. Deviation

Std. Error Mean

df

Sig. (2-tailed)

95% Confidence Interval of the


Difference
Lower

Pair 1

Marathon1 - Marathon2

22.128

21.486

3.440

15.163

Upper
29.093

6.432

38

.000

Results indicated that Marathon 1 (M = 226.82, SE = 9.80) took significantly longer


to run than Marathon 2 (M = 204.69, SE = 8.37), t(38) = 6.43, p < 0.001).

Вам также может понравиться