You are on page 1of 53

SPSS

SPSS
ONE-SAMPLE T-TEST
Janet is interested in determining if male adolescents
who do not engage in sports are more depressed than the
average male adolescent. To test this hypothesis, she
obtains Kansas University Depression Inventory (KUDI)
scores from 30 adolescent boys who have indicated on a
questionnaire that they do not engage in sports outside of
school gym classes. The KUDI is a depression measure
that was standardized to have a mean of 50 for male
adolescents. Janet conducts a one-sample t test on the
30 KUDI scores to determine if the population mean is
different from 50. The test value of 50 was chosen
because a value less that 50 implies more depression than
a typical male adolescent, a value greater than 50
implies more depression than a typical male adolescent,
and a value of 50 implies neither less not more
depression than a typical male adolescent. Janet’s SPSS
data file for analyzing the results contains 30 cases, one
for each adolescent, and one test variable, the KUDI
score.
Results Section

A one-sample t-test was conducted on the KUDI


scores to evaluate whether their mean was
significantly different from 50, the accepted mean
for male adolescents in general. The sample mean
of 54.63 (SD = 10.33) was significantly different
from 50, t(29) = 2.46, p = .020.
The results support the conclusion that young
male adolescents who do not engage in sports
outside of school are somewhat more depressed
than average.
SPSS
INDEPENDENT-SAMPLES T-TEST
Bart is interested in determining if individuals talk more when they are nervous. To evaluate his
hypothesis, he obtains 30 volunteers from undergraduate classes to participate in his experiment.
He randomly assigns 15 students to each of the two groups. Next, he schedules separate
appointments for every students.
When students in the first groups arrive for their appointments, they are told that in 10 minutes
they will be asked to describe how friendly they are by completing a true-false personality
measure. In contrast, students in the second group are told that in 10 minutes they will be asked
a set of questions by a panel of measurement specialists to determine how friendly they are. In
both groups, the conversations between the students and the experimenter are recorded for 10
minutes.
During this 10-minutes period, the experimenter makes prescribed comments if no conversation
occurs within a 1-minute time interval and responds to student questions with a brief response
such as yes or no. The percent of time that each student talks during the 10-minute period is
recorded.
Bart’s SPSS data file includes a grouping variable and a test variable for 30 cases. The grouping
variable distinguishes between the low stress group, whose members anticipate completing a
personality measure, and the high stress group, whose members anticipate questions from a
panel of measurement specialists. The test, or dependent, variable is the percent of time that
each student talks during the 10-minute period.
Results Section
An independent-sample t test was conducted to evaluate the hypothesis that students talk more
under a high-stress condition (anticipating questions from a panel of measurement specialists) as
opposed to a low-stress condition (anticipating taking a personality measure). The test was
significant, t(28) = 2.43, p = .022, but the result were counter to the research hypothesis. Students
in the high-stress condition (M = 22.07, SD = 27.14) on the average talked less than those in the
low-stress condition (M = 45.20, SD = 24.97).
SPSS
ONE-WAY ANALYSIS OF VARIANCE (ANOVA)
Dana wishes to assess whether vitamin C is effective in the treatment of colds. To
evaluate her hypothesis, she decides to conduct a 2-year experimental study.
She obtains 30 volunteers from undergraduate classes to participate. She randomly
assigns an equal number of students to three groups: placebo (group1), low doses of
vitamin C (group 2), and high doses of vitamin C (group 3).
In the first and second years of the study, students in all three groups are monitored to
assess the number of days that they have cold symptoms. During the first year,
students do not take any pills.
In the second year of the study, students take pills that contain one of the following:
no active ingredients (group 1), low doses of vitamin C (group 2), or high doses of
vitamin C (group 3).
Dana’s SPSS data file includes 30 cases and two variables: a factor distinguishing among
the three treatment groups and a dependent variable, the difference in the number of
days with cold symptoms in the first year versus the number of days with cold
symptoms in the second year.
Vitamin C Treatment Low doses of vitamin C High doses of vitamin C Placebo
difference in the number
of days with cold for first
B B A
year and second year -2.10 -2.00 3.50
Results Section
A one-way analysis of variance was conducted to evaluate the relationship
between vitamin C and the change in the number of days with cold symptoms from
the first year to the second year of the study. The independent variable, the
vitamin C factor, included three levels: placebo, low doses of vitamin C, and high
doses of vitamin C. The dependent variable was the change in the number of fays
when cold symptoms were exhibited from the first year to the second year.
The ANOVA was significant, F(2,27) = 4.84, p = .016. Follow-up test of Scheffé
revealed that there were significant differences in the means between group that
received a low and high dose of vitamin C and the placebo group, The group that
received a low and high dose of vitamin C showed a greater decrease in number of
days with cold symptoms in comparison to the placebo group.
normally distributed
Generally, when n30, normal distribution can be assumed, but if size of n is less than
30 the data should be tested by Kolmogorov-Smirnov test or Shapiro-Wilk (use when
size of n is less than or equal to50), whether it is normally distributed.
 
Assumption for Kolmogorov-Smirnov and Shapiro-Wilk
 
Ho: The variable is normally distributed
H1: The variable is not normally distributed
 
SPSS: Analyze  Descriptive Statistic  Explore Plot  Normality plots with tests
 
 Result
 
Level of Significant is more than .05  fail to reject Ho (the variable is normally
distributed)
NB. Shapiro-Wilk should be read only if the sample sizes are less than or equal to 50.
SPSS
NORMALLY DISTRIBUTED
Level of Significant is more than .05  fail to reject Ho (the variable is normally distributed)
NB. Shapiro-Wilk should be read only if the sample sizes are less than or equal to 50.

Report: The assumption of Normality was analyzed using two tests of significance: the Kolmogorov-Smirnov test and the
Shapiro-Wilk test. The Kolmogorov-Smirnov test showed a probability coefficient of 0.2 for each group since this value is
greater than 0.05, the Kolmogorov-Smirnov test did not reject the null hypothesis that the scores for each group is normally
distributed.
The Shapiro-Wilk test showed a probability coefficient of 0.49 for method 1, .667 for method 2, and 0.437 for method 3. In all
three cases the coefficient is greater than 0.05. Therefore, the Shapiro- Wilk test did not reject the null hypothesis that the
scores for each group are normally. The results for both the Kolmogorov-Smirnov test and Shapiro-Wilk test provide support
for the assumption of normality.
SPSS
CORRELATION & REGRESSION
Simple regression and Correlation Analysis is used for studying the relationship between 2 or more variables
Although the two terms are often interchangeably, there is a difference between the purpose and the usage of the terms
Correlation is a measure of the closeness of linear relationship between two or more interval scale variables
Regression is used to derive an equation that relates the single dependent variable with 1 or more predictor variables,
when there is 1 predictor variable (simple regression), whereas there are 2 or more predictor variables (multiple
regression)
the equation will be in the form of
y = a+bx
a = slope of the regression line
b = y intercept (y when x = 0)
Assumption for Regression Analysis
Mean of error term is zero (always true, do not need to test)
Variance of the error is constant (plot graph between 2 with X,Y)
The error terms are uncorrelated and the observations have been drawn independently. (look if Durbin-Watson
value
 close to 0 = high positive correlation
 close to 4 = high negative correlation
 around 2 = uncorrelated
The error term is normally distributed (K-S)
Additional assumption for Multiple Regression
Xi and Xj (independent variables) have been drawn independently. (look at one or more of the following values
Eigenvalue close to zero = dependent relationship
Low Tolerance value = high relationship
High VIF value = high relationship
Condition index more than 20 = existence of relationship
Salary (Baht) Age (yr.) HR test scores

10,000 21 150
13,000 23 130
Example: Regression Analysis
14,000 24 120
SPSS: Analyze  Regression  Linear 17,000 27 140
18,000 28 140
Select Method (read text)
18,500 28 150
18,500 27 135
STAT  Select Durbin-Watson 19,000 26 160
19,500 29 140
Select Salary as Dependent  Age as Independent
20,000 27 125
10,000 21 150
13,000 23 130
14,000 24 120
17,000 27 140
18,000 28 140
18,500 28 150
18,500 27 135
19,000 26 160
19,500 29 140
20,000 27 125
Variables Entered/Removed b

Variables Variables
Model Entered Removed Method
1 AGEa . Enter
a. All requested variables entered.
b. Dependent Variable: INCOME

Model Summary b

Adjusted R Std. Error of


Model R R Square Square the Estimate Durbin-Watson
1 .928a .861 .853 1231.18 2.013
a. Predictors: (Constant), AGE
b. Dependent Variable: INCOME
ANOVAb

Model Sum of Squares df Mean Square F Sig.


1 Regression 168965517.241 1 168965517.2 111.469 .000a
Residual 27284482.759 18 1515804.598
Total 196250000.000 19
a. Predictors: (Constant), AGE
b. Dependent Variable: INCOME

Coefficients a

Standardize
Unstandardized d
Coefficients Coefficients
Model B Std. Error Beta t Sig.
1 (Constant) -14629.310 2984.840 -4.901 .000
AGE 1206.897 114.312 .928 10.558 .000
a. Dependent Variable: INCOME
Residuals Statistics a

Minimum Maximum Mean Std. Deviation N


Predicted Value 10715.52 20370.69 16750.00 2982.10 20
Residual -1163.79 2250.00 -1.82E-13 1198.34 20
Std. Predicted Value -2.024 1.214 .000 1.000 20
Std. Residual -.945 1.828 .000 .973 20
a. Dependent Variable: INCOME
Result…
Durbin 2.013 close to 2 mean uncorrelated,
ANOVA table
Ho: No linear relationship
H1: Linear relationship
Sig. Reject Ho  linear relationship
Y= a+bx
Y = salary; a = constant value ( -14,629.3) b= independent variable (1,206.89)
R square = .861

Results
A (multiple) linear regression analysis was conducted to evaluate the prediction of the
average age and salary level for graduates. The results of the analysis indicated that age
accounted for a significant amount of salary for graduates, R 2 = .86 F(1,18) = 111.47, p < .01,
indicating that the older graduate tend to receive higher level of salary.
The Impact of Cultural Value Orientation on Customer Perceptions of Post-Recovery Service Satisfaction
Paul G. Patterson, Elizabeth Cowley, Kriengsin Prasongsukarn, IJMR (Forthcoming)
SPSS
MEASUREMENT RELIABILITY & VALIDITY
Example: Cronbach alpha (α)
Example: Cronbach alpha (α)
Satisfaction with service recovery was measured using the scales
adapted from Oliver and Swan (1989). Overall satisfaction with the
service recovery was measured with three seven-point scales
anchored with “very dissatisfied” and “very satisfied”; “did a poor
job” and “did a good job”; and “happy” and “unhappy” (α = .89).
Exploratory Factor Analysis (EFA)
The perception of justice was measured using scales
adapted from Smith, Bolton, and Wagner (1999). The
perception of justice has three dimensions: distribution
justice, procedural justice, and interactional justice
(see Table 5.6 for the exact wording of the questions).
Distributive justice was measured with 4 items (α
= .89). Procedural justice was measured with two
items (r = .71). Interactional justice was measure with
four items (α = .82).
Exploratory Factor Analysis (EFA)
The exploratory factor analysis of perception of justice
scales resulted in a clean solution, all items loaded
highly on the appropriate factors and no item loaded
on more than one factor. All 10 items loading on the
dimensions as reported in Smith, Bolton, and Wagner
(1999), thus indicating convergent and discriminant
validity. See Table 5.7 for reports of factor loadings
and variance explained. All of the distributions were
approximately normal. The ten items were then
subjected to a confirmatory factor analysis.
SPSS
CHI-SQUARE
Statistic Associated with Cross-Tab
Chi-square (X2) test
Test the null hypothesis, Ho is that there is no association between the variables
assists us in determining whether a systematic association exists between the two variables.
can also be used in goodness-of-fit tests to determine whether certain models fit the observed data.
should be estimated only on counts of data. When the data are in % form, they should be converted
to counts or number.
particularly useful in test involving nominal data but also can be used for higher scales.
Use of Chi-square in SPSS
1.Pearson Chi-square
1. In 2 x 2 tables, each cell should contains at least 5 sample sizes
2. In larger tables, the minimum sample sizes can be 1 if the number of cells that
are less than 5 are not more than 20% of all cells
2.Linear-by-Linear Association
1. Tests for linear relationships when both variables are collected in INTERVAL
scales.
3.Likelihood Ratio Chi-square
1. Suitable for multiple-ways table and when data are collected in an ORDINAL
manner.

SPSS: Analyze  Descriptive Statistics  Crosstabs  Statistic Chi-square


A two-way contingency table analysis was conducted to assess
the relationship between taking a tutorial lesson (taking and not
taking a tutorial class) and students’ gender (male and female).
These variables were not significantly related, Pearson 2(1,300) =
1.63, p>.05.
Drawbacks
The larger size of sample will increase the value of Chi-square
Lowest value for Chi-square is equal to zero and it have no highest value, so
we can tell if there is relationship, but we can not specific its strength.
Chi-square has only positive (+) value, so we can not specify direction of the
relationship.
Determining the strength for data collected by NOMINAL scale
The following statistic tests are available in the SPSS program to determine the
strength of the relationships between variables measured on a NOMINAL scale.
(For nominal data, direction is not exist)
1. Contingency Coefficient (can be used for contingency tables of any size)
2. Phi (use with 2 x 2 table)& Gramer’s V Coefficient (can use with tables larger than
2 x 2 table)
3. Lambda Coefficient
1. Asymmetric Lambda – measures the percentage improvement in predicting the
value of the dependent variable, given the value of the independent variable.
2. Symmetric Lambda – does not make an assumption about which variable is
dependent. It measures the overall improvement when prediction is done in
both directions.
4. Uncertainty Coefficient – developed from likelihood ratio
NB. The strength of association range from 0<x<1, o being no association, and 1 being
perfect association
Determining the strength and direction for data collected by ORDINAL scale
The following statistic tests have been introduced in SPSS program to determine the
strength and direction of data collected by ORDINAL scale.
1. Gamma – does not make an adjustment for table size, generally has a higher
numerical value than either tua-b or tua-c (see table below)
2. Sommer’s d – similar to Asymmetric lambda, measure the % improvement in
predicting the value of the dependent variable
3. Kendall’s tau-b – appropriate with square table, in which the number of rows
and columns are equal
4. Kendall’s tau-c – can be used with rectangular table in which the number of
rows is different than the number of columns
NB. Value range from –1 to +1; the value that is closer to –1 or +1 means the
stronger the strength of the relationship and the positive/negative sign represents
the direction of such relationship.
Note
Data used in these slides were taken from “Using SPSS For Windows: Analyzing and
Understanding data” 2nd Ed, Green S.B., Salkind, N.J., and Akey, M.T.

You might also like