Transcript kin260_lec8
SPSS Part 2
Kin 260
Jackie Kiwata
Overview
Review
Comparing Sets of Data
Correlation
T-Tests
Overview in depth
Once data is organized, we generally analyze
the data by
Evaluating raw scores
-OR Comparing sets of data
Last time we evaluated raw scores
For instance, we calculated percentiles
Today, we’ll compare sets of data using SPSS
Statistical Significance Tests
Correlation
Basic idea: Use to determine if a relationship
exists between variables
T-tests
Basic idea: Use to determine if the means of 2
samples are statistically the same or different
Correlation Review
Indicates the extent to which two variables are related.
The technique used to measure this is Pearson’s correlation
coefficient, r
The closer r is to 1 or -1, the stronger the relationship
But Pearson’s correlation coefficient does not indicate
causation
Not correct to say X causes Y
The strength of the relationship is classified using:
.9 or greater
strong
.8 - .9
moderately strong
.7 - .8
moderate
.5 - .7
low
< .5
no relationship
Correlation: SPSS example
Research question: Can a person’s height predict
their vertical jump height?
Height (in)
11.5
17.5
19.5
11
23.5
23
Vertical Jump Height
(in)
64
66
69
68
67
68
Correlation in SPSS
Analyze > Correlate > Bivariate …
A word about Tails…
Two-tailed test
Use if prior research or logical reasoning does not
clearly indicate a significant difference between
the mean values should be expected
Will use most of the time
One-tailed test
Use if the direction (+ or -) of the difference
between the means is well established before
data collection
Sample Correlation Output
De scri ptive Statistics
VJ
Height
Mean
17.667
67.00
St d. Deviat ion
5.4467
1.789
N
6
6
R value
Corre lations
VJ
VJ
Height
Pearson Correlation
Sig. (2-tailed)
N
Pearson Correlation
Sig. (2-tailed)
N
1
6
.431
.393
6
Height
.431
.393
6
1
6
Significance
Significance
The probability that:
a result is not likely to be due to chance alone
a result is correct
Expressed as a p value (probability value)
In research, significance is set before
collecting data
Levels of Confidence &
Probability of Error
In general, statistical significance is reported
at one of three levels:
If p=0.01, 99% confident the results are correct
and 1% incorrect
If p=0.05, 95% confident the results are correct
and 5% incorrect
If p=0.10, 90% confident the results are correct
and 10% incorrect
If p>0.10, don’t report at all, because not
statistically significant
Significance Example
Suppose SPSS gives p = 0.02
At which level is this value significant?
To answer this question, compare p value to
each level of confidence
If p is less than a given level, try the next
P < 0.10
P < 0.05
P < 0.01
Significance example, con’t.
Report
as NS.
No
Is p < 0.10 ?
Y
Report
as
p<.10
No
Is p < 0.05 ?
Y
Report
as
p<.05
No
Is p < 0.01 ?
Yes
Report
as
p<.01
For p=0.02,
We say p is significant at p<0.05 or at the 95% LOC
Correlation Significance
Example
R=0.75 but p=0.35
Obtained a moderate correlation, but we are
only 65% certain results are correct and due
to chance
P=0.35 is not statistically significant, so we
can conclude results not due to chance
Possible reasons
Sample is not representative of population
Data has been tampered with in some way
T-test
Use to compare one sample mean to the
population mean
-OR Use to compare two independent, unrelated
samples drawn from same population
Tells us if the two means are statistically
different
Like correlation, the T test does not
indicate causation
Types of T tests
Independent Samples
Run this test if just need to compare the means
between two groups
e.g. Obtained VO2max scores from Kin 260 Sec 1
and Sec 3. Want to compare the means between
the two sections
Paired Samples
Run this test if research design required a pre and
post test and same subjects were tested twice
e.g. All Kin 260 students took a VO2max test
before beginning a conditioning program, then took
another VO2max test 6 weeks later.
Ex – Independent Sample T Test
A biomechanics student gave a sit-and-reach test to 10 male
and 10 female undergrads.
The following measurements in cm were obtained:
Males
Females
19.1
20.4
17.2
25.0
20.1
26.9
18.2
27.1
16.5
28.3
16.9
22.2
21.2
23.8
19.7
24.8
15.9
25.4
16.0
19.0
T-Test Example con’t.
Male mean = 18.1
Female mean = 24.3
Appears females are more flexible at the hip
than males
But is this difference statistically significant?
T-test: SPSS
Define variables in Variable View
1.
-
2.
3.
4.
5.
ONE BIG DIFFERENCE: Need to create a
“grouping” variable using Value Labels
Enter data in Data View
Analyze > Compare Means > Independent
Samples T test…
Add variables
Define Groups
T-test: SPSS con’t.
T test: SPSS
T-test: SPSS Output
•SPSS reports highly significant values as 0.00
- take this to be p<0.01
• So t = 5.573 and p<0.01
T test - Results
If p<0.10, get to say difference between the
means is statistically significant
report t value and level of confidence
But remember t value does not tell us why the
means are different
If p>0.10, report as NS (not significant)
One or both of these samples were not randomly
drawn, OR
Some factor has affected these samples, causing
them to be different from the original population
Ex – Paired T test
•A running coach gave a 1.5 mile running test to
her athletes before and after a 3 wk
cardiovascular training program
•Did the training program improve running time?
Pre (min)
Post (min)
9.5
9.1
12.2
12.0
12.8
12.6
10.2
10.2
10.8
10.9
9.5
9.4
Paired t test steps similar to
independent t test
Define variables in Variable View
1.
2.
3.
4.
Do not need to use Value Labels
Set up variables in same manner as
correlation
Enter data in Data View
Analyze > Compare Means > Paired
Samples T test…
Add variables
Correlation vs. T-test
When and how to use each?
Correlation
Does x data predict y
data?
Can y data be predicted
from x data?
Graph in Excel, analyze
data in SPSS.
T test
Compare the means
between the two groups.
Compare the pre and post
tests.
Analyze data in SPSS.
More Information
SPSS:
http://www.calstatela.edu/its/docs/pdf/SPSS1
4Part2.pdf
T-tests:
http://en.wikipedia.org/wiki/Student's_t-test