Information Technology Reference
In-Depth Information
GAFscore
Cumulative
Percent
Frequency
Percent
Valid Percent
Valid
35
39
45
50
55
60
68
69
70
71
72
Total
1
6.7
6.7
6.7
1
6.7
6.7
13.3
1
6.7
6.7
20.0
2
13.3
13.3
33.3
2
13.3
13.3
46.7
1
6.7
6.7
53.3
1
6.7
6.7
60.0
1
6.7
6.7
66.7
2
13.3
13.3
80.0
1
6.7
6.7
86.7
2
13.3
13.3
100.0
15
100.0
100.0
Figure 5.10
Frequency table for GAFscore.
5.5.2 SPSS MISSING VALUES ANALYSIS
Before considering issues of normality, and homogeneity of variance,
researchers should first address the companion issues of missing values
and outliers. Left unattended, either of these issues could distort the shape
of a distribution.
The present example data set contained no missing values for the
variables. Had there been one or more missing values for either of the
variables, we could have applied one of two strategies: (a) Use the SPSS
or SAS listwise deletion default, where the program computes statistics
(e.g., means) by omitting cases that have missing values on a particular
variable; (b) try to estimate or
impute
the missing value with a statisti-
cal replacement value. SPSS offers the
Missing Values Analysis (MVA)
module as a special add-on to the SPSS system. SPSS also offers a mean
substitution imputation procedure through the
Transform
facility. Like-
wise, SAS offers the
MI (Multiple Imputation)
program for imputing
missing values. A full discussion of analyzing missing values is beyond the
scope of a text on ANOVA like the present topic. As a brief introduction
we recommend parts of chapters 3A and 3B of Meyers et al. (2006), and for
a more complete discussion see Allison (2002) or McKnight, McKnight,
Sidani, and Figueredo (2007).
5.5.3 SPSS OUTLIERS ANALYSIS
Recall that outliers refer to cases with extreme values on a particular
variable that does not represent the population under study. To assess for
outliers on our continuous dependent variable (GAFscore) across levels
of our categorical independent variable (therapy), we begin by clicking
Analyze
➜
Descriptive Statistics
➜
Explore
, which opens the
Explore
dialog window (see Figure 5.11).
