~Drowning in Data~
SPSS Data Analysis
3/26/12
Sumiko Takayanagi, Ph.D.
Sr. Statistician
UCLA School of Nursing
1
Today’s Presentation

SPSS Environment
 Review of SPSS Basics

Inferential Statistics in SPSS
 Independent t-test
 Two-Way Analysis of Variance
 Multiple Regression

Conclusion

References
2
Features of SPSS




Originally developed for the people in Social
Science Areas, therefore, no heavy
programming background required
Designed as User Friendly and has Pull
Down Menus to Execute Statistical
Commands
Ability to do Data Management &
Manipulations
Ability to Store Programs & Produce
Reports/Graphs
3
SPSS Program Flow
Outside
Data
Source
Raw
Data
SPSS
Data
File
Data
Modification/
Transformation
Data Analysis
Pull-Down
Menu
OR
Syntax
Menu
(Data Steps)
(Analysis Steps)
4
Data View Window - Data Entry Site
(Columns=Variables, Rows=Cases)
Help Menu
Pull-down Menu bar
Tool bar
Information bar
Title bar
Variable
Names
Data View window
Active cell
Action bar
5
Variable View Window
Data Definition Site
64
Characters
Max, No
space
Between
Beg letter,
@, #, or $
Numeric,
String, &
Others
Length
# of
Decimals
Variable
Description
Value
Code
Description
Click here to see this view
Missing
value
Description
6
Before we see
Examples…
OK
VS.
Paste
buttons
<Output File>
1. OK - results/action
will be executed
7
1. Hit Paste to obtain
Syntax Window
2. Run Syntax
to obtain the
results in the
Output
Window
<Syntax File>
8
Example - School Data
Raw Data



Subject 1

Subject #

Female

Intensive

Reading

Math
(1)
(1)
(1)
(90)
(67)
Subject 2

Subject #

Female

Moderate

Reading

Math
(2)
(1)
(2)
(72)
(46)
Subject 3

Subject #

Male

Basic

Reading

Math
(3)
(0)
(3)
(41)
(73)
9
School Data
Variable View
Variable View Activated
10
School Data
Completed Dataset – Data View
11
School Data
Completed Dataset – Variable View
12
Importing Excel Data file to SPSS
1. Open the SPSS Data file
2. Go to File Menu
3. Click “Read Text Data”
4. Click Files of type to Excel
& choose Excel file
5. Hit Open
6. Check Worksheet #,
Variable on the 1st row, &
Hit OK
13
School Data
Completed Dataset – Data View
14
Click to Obtain
Data File Information
15
Variable Information
16
Value Code Information
17
Basic Statistical Methods



Independent t-test
Two-Way ANOVA
Multiple
Regression
18
Independent t-test
– Is there a significant difference between 2 groups?
Assumptions 1. Normality
2. Variance
Equality
# of Variables
Characteristics
Dependent = 1
Continuous
Independent = 1
Categorical
2-levels
3.
Independence
School Data
N=100
Math Score
Range of 0-100
Gender
19
How to calculate t-value?
t-value=
Mean Difference
Group Variability
20
t-test
Medium
Variability
High
Variability
Low
Variability
21
Independent t-test
1. Go to Analyze.
2. Choose
Compare Means.
3. Choose
Independent
Samples t Test.
22
t-test
1. Choose Dependent
& Independent Variables.
23
Descriptives &
Analysis
Independent Variable
Dependent Variable
Variance Equality Test
t=
Z1 – Z2
SD12 + SD22
N1
N2
=
t - statistics
63.20 – 54.10
(13.914)2 +(13.064)2
41
=
t=
Mean Diff
Std. Error Diff
9.093 = 3.295
2.760
59
24
Conclusion &
Chart

There is a
significant
difference
in math
ability
between
males and
females.
Males
performed
better than
females.
25
Factorial ANOVA
– Is there any main or the interaction effects?
Assumptions 1. Normality
2. Variance
Equality
# of Variables
Characteristics
Dependent = 1
Continuous
Independent >1
Categorical2 or more levels
3.
Independence
School Data
N=100
Math Score
0-100
Gender
Program Type
26
2 x 3 Factorial ANOVA
Design Diagram
Gender
Male
Female
Mild
56, 86, 70, 69,
…..
55, 72, 67, 48,
…..
Moderate
86, 59, 67, 80,
…..
63, 78, 55, 46,
…..
Intensive
89, 92, 86, 71,
…..
72, 76, 54, 56,
…..
Program
Math Test Scores
27
2-Way Factorial ANOVA
1.Go to General
Linear Model &
choose Univariate.
2. Choose One Dependent
& Two Independent
Variables.
28
Factorial ANOVA (2x3)
Descriptives
1. Freq of IV and Raw Means
2. Equality of Variance Test
29
Factorial ANOVA
Main Analysis
Main Effects &
Interaction

Results:


Main effect – Sig. difference in gender and in program type.
Interaction – Sig. interaction between gender and program
type.
30
Factorial ANOVA
Multiple
Comparison
Which levels
are actually
Different ??
Scheffe & LSD
Methods
Sig. Different
level
31
Factorial ANOVA
Conclusion
Significant Effects
 Males performed better
than females.
 Students in the Intensive
program performed
better than in the Mild
program.
 Males in the Intensive
program performed
better than in other
programs, but no
performance difference
in females.
32
Multiple Regression
– Which IVs can predict the DV and to estimate the effects of
these variables on DV?
Assumptions
1. Normality
# of Variables
2. Variance
Equality
3.
4. Linear
Independence Relationship
Characteristics
Health Survey
Data
N=100
Dependent =1
Continuous
Independent > 1
Continuous or
Dichotomous (0
or 1) Variables
LDL Value
0-200
HT, WT, BMI, &
Exercise
33
Multiple Regression Diagram
HT
DV
WT
LDL
IV
BMI
Exercise
All 4 IVs are predicting LDL
34
Health Survey Data of N=100
35
Multiple Regression
1.Choose Regression
2. Choose Linear Regression
36
2. Choose Statistics you need.
1. Choose DV, IV, & Method.
3. Choose Residual Plots.
37
Descriptives
& Correlation
Tables
Descriptive
Stats.
Correlation
Coefficients &
corresponding
p-values.
38
Main Analysis
R2=how much of the variability in the outcome is accounted
for by the predictors (regression sum of squared/total sum of squares)
Adj. R Sq=Adj for the # of
Parameters in the model
R=r between pred and
observ value of the DV
Global test to
see if any
coefficient is
different from
“0”
B=Reg Coefficient
t & Sig=IV
predictability
Partial/Part
Correlations
Tolerance
&VIF
Beta=Stdized. Reg
Coefficient.
Something is Wrong
if Beta >1!!
39
Residual
Analysis
Residual Normality
Linearity and
Equal Variance & residual independence
40
Conclusion




Multiple Regression
IVs explain about 40% of
the variability of LDL
level.
The significant
predictors of LDL were
BMI and Hrs of
Exercise.
The collinearity statistics
didn’t show exceptionally
large multicollinearity
among predictors.
Assumptions of residual
normality and equal
variance were met.
41
Key Concepts
Statistical Models depend on the theory
and data. Choose your model wisely to see
if it can answer your research questions.
 Check Assumptions. Model conclusions
may not be valid unless the assumptions
were met. If not, use appropriate
corrections, do data transformations, or
even use other statistical methods.

42
Conclusions

Statistical judgments come into
our daily lives. Statistics are more
than mathematical calculations or
scientific research, but they are the
way of logical thinking…
Thank you
43