Basic and Advanced Statistical Tests

Writing Results Sections and Creating

Tables and Figures 1st Edition Amanda
Ross
Visit to download the full and correct content document:
https://textbookfull.com/product/basic-and-advanced-statistical-tests-writing-results-se
ctions-and-creating-tables-and-figures-1st-edition-amanda-ross/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Basic Chemical Concepts and Tables 1st Edition Steven

L. Hoenig

https://textbookfull.com/product/basic-chemical-concepts-and-
tables-1st-edition-steven-l-hoenig/

Statistical analysis of contingency tables 1st Edition

Fagerland

https://textbookfull.com/product/statistical-analysis-of-
contingency-tables-1st-edition-fagerland/

Basic Chemical Concepts and Tables, 2nd Edition Steven

L. Hoenig

https://textbookfull.com/product/basic-chemical-concepts-and-
tables-2nd-edition-steven-l-hoenig/

Spock Up and Running Writing Expressive Tests in Java

and Groovy 1st Edition Fletcher

https://textbookfull.com/product/spock-up-and-running-writing-
expressive-tests-in-java-and-groovy-1st-edition-fletcher/
Simple Statistical Tests for Geography 1st Edition
Danny Mccarroll

https://textbookfull.com/product/simple-statistical-tests-for-
geography-1st-edition-danny-mccarroll/

Statistical Methods An Introduction to Basic

Statistical Concepts and Analysis 2nd Edition Cheryl
Ann Willard

https://textbookfull.com/product/statistical-methods-an-
introduction-to-basic-statistical-concepts-and-analysis-2nd-
edition-cheryl-ann-willard/

Decode Mathematics Advanced Topic Tests 1st Edition

Thushan Hettige

https://textbookfull.com/product/decode-mathematics-advanced-
topic-tests-1st-edition-thushan-hettige/

Linear and Nonlinear Circuits Basic and Advanced

Concepts Mauro Parodi

https://textbookfull.com/product/linear-and-nonlinear-circuits-
basic-and-advanced-concepts-mauro-parodi/

Analysis in Nutrition Research: Principles of

Statistical Methodology and Interpretation of the
Results 1st Edition George Pounis

https://textbookfull.com/product/analysis-in-nutrition-research-
principles-of-statistical-methodology-and-interpretation-of-the-
results-1st-edition-george-pounis/
Basic and Advanced Statistical Tests
Basic and Advanced Statistical Tests
Writing Results Sections and Creating Tables and Figures

Amanda Ross
A. A. Ross Consulting and Research, USA
and
Victor L. Willson
Texas A&M University, USA
A C.I.P. record for this book is available from the Library of Congress.

ISBN: 978-94-6351-084-4 (paperback)

ISBN: 978-94-6351-085-1 (hardback)
ISBN: 978-94-6351-086-8 (e-book)

Published by: Sense Publishers,

P.O. Box 21858,
3001 AW Rotterdam,
The Netherlands
https://www.sensepublishers.com/

For many of our analyses, the authors used the BASC data set, Results of Dataset
Analysis from the Behavior Assessment System for Children, Second Edition (BASC-2).
Copyright © 2004 NCS Pearson, Inc. Analysis included with permission. All rights
reserved.

Printed on acid-free paper

All Rights Reserved © 2017 Sense Publishers

No part of this work may be reproduced, stored in a retrieval system, or transmitted

in any form or by any means, electronic, mechanical, photocopying, microfilming,
recording or otherwise, without written permission from the Publisher, with the
exception of any material supplied specifically for the purpose of being entered and
executed on a computer system, for exclusive use by the purchaser of the work.
 TABLE OF CONTENTS

Introduction ix

Part I: Basic Statistical Tests

Chapter 1: Descriptive Statistics 3

Brief Description 3
Writing a Results Section 5
Results Section Write-up 6

Chapter 2: One-Sample T-Test 9

Brief Description 9
Writing a Results Section 11
Results Section Write-up 12

Chapter 3: Independent Samples T-Test 13

Brief Description 13
Writing a Results Section 15
Results Section Write-up 16

Chapter 4: Paired Samples T-Test 17

Brief Description 17
Writing a Results Section 18
Results Section Write-up 19

Chapter 5: One-Way ANOVA 21

Brief Description 21
Writing a Results Section 23
Results Section Write-up 24

Chapter 6: Factorial ANOVA 25

Brief Description 25
Writing a Results Section 28
Results Section Write-up 29

v
TABLE OF CONTENTS

Chapter 7: MANOVA31
Brief Description 31
Writing a Results Section 35
Results Section Write-up 37

Chapter 8: Linear Regression 39

Brief Description 39
Writing a Results Section 43
Results Section Write-up 46

Chapter 9: Multiple Regression 49

Brief Description 49
Writing a Results Section 57
Results Section Write-up 58

Part II: Advanced Statistical Tests

Chapter 10: Hierarchical Multiple Regression Analysis Using at Least

Two Sets of Variables (in Two Blocks) 61
Brief Description 61
Writing a Results Section 65
Results Section Write-up 68
Results Section Write-up 73

Chapter 11: Multiple Regression with Two Continuous Predictors

and the Interactions between Them: Data Centered 75
Brief Description 75
Writing a Results Section 81
Results Section Write-up and Table 86
Chapter 12: Two or Three Factor Crossed ANOVA87
Brief Description 87
Post Hoc Tests 92
Writing a Results Section 102
Post Hoc Tests 108
Post Hoc Tests 114
Results Section Write-up 119

Chapter 13: ATI Analysis (ANCOVA with Interaction) 121

Brief Description 121
Writing a Results Section 131
Results Section Write-up 143

vi
 TABLE OF CONTENTS

Chapter 14: Multiple Regression with at Least Three Predictors: Use NORM
to Impute Missing Data 145
Brief Description 145
Writing a Results Section 149
Results Section Write-up 159

Chapter 15: Structural Equation Model Analysis with at Least One

Indirect Path 161
Brief Description 161
Model Fit Summary 165
Writing a Results Section 167
Results Section Write-up 179

Chapter 16: MANOVA and Follow-up Discriminant Analysis with at

Least Three Groups and Four Independent Variables 183
Brief Description 183
Writing a Results Section 205
Post Hoc Tests 212
Results Section Write-up219

vii
 INTRODUCTION

Many researchers have difficulty knowing how to properly write a results section for
a scholarly work. They may comb through countless journals, trying to find tables
and/or figures that match the statistical test used in their design. Statistics courses do
a fine job of teaching graduate students how to run various basic and advanced tests.
However, students are often left to their own devices to determine how to write the
results section and create the appropriate tables and/or figures in APA format.
This book presents a concise look at basic and advanced statistical tests, providing
a brief description of each and examples of research problems that yield themselves
to each particular test. Revealing the core strength of each chapter, a sample scenario
and results section write-up are provided. Depending upon the test and need, a
sample table and/or figure may be provided.
This book serves as a reference manual for all professionals in the psychology,
sociology and education fields. Actually, this book can assist professional researchers
in any field, including medical research. By providing specific examples of situations
aligned with each type of test, the reader can become adept at understanding what
different variables (nominal, ordinal, interval, and ratio) look like in real-world
problem situations.
While other books provide an in-depth examination of each test, how to run each
test, and ways to interpret results, this book provides the missing piece in the literature
out there. Graduate students, teachers, faculty, social workers, psychologists, and
other professional researchers now have access to a guide that includes the name of
the test, a brief description (without all of the technical jargon), examples, a sample
scenario, a sample results section write-up, and accompanying tables and/or figures!
When having difficulty knowing which information to extract from the output,
simply turn to each results section. If you aren’t sure which information to place in
the write-up and which to place in a table, simply look at the examples!
We do hope this book/reference manual is helpful to you. As a graduate student
researcher, this sort of book would have helped me immensely throughout my
research endeavors. It provides us great pleasure to share this reference with you.
Good luck with all your current and future research studies!

For many of our analyses, we used the BASC data set, Results of Dataset analysis
from the Behavior Assessment System for Children, Second Edition (BASC-2).
Copyright © 2004 NCS Pearson, Inc. Analysis included with permission. All rights
reserved.

ix
 PART I
BASIC STATISTICAL TESTS
 CHAPTER 1

DESCRIPTIVE STATISTICS

BRIEF DESCRIPTION

The field of statistics can be divided into the categories of descriptive statistics and
inferential statistics. Descriptive statistics are used to characterize a sample and
include computations that do not require any inference about a population, such
as frequencies, percentages, median, mean, mode, standard deviation, z-scores,
variance, range, and interquartile range. Such statistics prove very useful when
sample sizes are not large enough to warrant other tests.
Frequencies and percentages are low-level quantitative approaches that simply report
either a count or a ratio. They can be easily represented, using bar graphs or histograms.
Scores are often grouped, such as into fourths, termed quartiles, or tenths, termed deciles.
Measures of the center of a distribution of data include the median, mean, and
mode. The median represents the middle value (or average of the two central values)
of the data, when data are arranged in ascending order. The median is not affected
by extreme scores and should be used as the measure of center when data includes
extreme outliers or is not approximately normally distributed. The mean is equal to
the ratio of the sum of all scores to the total number of scores. The mean is affected
by extreme scores. The mean is appropriate when data is normally or symmetrically
distributed. The mode is simply the data point that occurs most often in a data
set. The mode is often considered the least useful measure of center, due to the
possibility that the most frequent score occurs well away from the center and is not
very representative of most scores.
Measures of deviation or spread of scores include standard deviation, variance,
range, and interquartile range. The standard deviation of a set of scores is related to
the typical deviation of a score from the mean, somewhat analogous to an average
distance from the mean. The variance of a set of scores is simply the square of
the standard deviation. The range is equal to the difference between the maximum
and minimum score. The interquartile range is the difference between the score
representing the third quartile, and the score representing the first quartile. The first
quartile score represents the median of the lower half of the scores, while the third
quartile score represents the median of the upper half of the scores.
Z-scores are standardized scores, with a mean of 0 and standard deviation of 1.
Z-scores may be useful when comparing an individual score to a population mean,
or when comparing a sample mean to a population mean, when given the population
standard deviation. They can be used with any set of scores, often to interpret or
compare two scores computed using different metrics. For example, a student might

3
CHAPTER 1

have a mean of 88 on one test and 83 on another. However, the mean for the first test
was 74, standard deviation 12, while the mean of the second test was 66, standard
deviation 15. The respective z-scores are 1.17 and 1.13, virtually identical in relation
to their means, even though it was not obvious from the original data.
Note. For a normal distribution in a population, a z-distribution based on the
normal curve, may be used whenever the population standard deviation is known.
A z-score may be interpreted as the number of standard deviations a value falls
above or below a mean. For example, suppose a student’s score on a statistics exam
shows a z-score of −1.5. This z-score reveals that the student’s score is 1.5 standard
deviations below the class mean for the exam.
X −m
The formula, z = , may be used to compare an individual score to
s
a population mean, where X represents the individual score, μ represents the
population mean, and σ represents the population standard deviation. Note. If only
a sample mean and population standard deviation are available, the sample mean
may be substituted for the population mean, in the formula shown above. However,
if the population standard deviation is not known, and only the sample mean and
sample standard deviation are available, a t-distribution should be used, instead. This
distribution is a bit wider and flatter than the normal or “bell” curve since it accounts
for the uncertainty in the standard deviation.
X −m
When comparing a sample mean to a population mean, the formula, z = , may
s
n
be used, where X represents the sample mean and n represents the sample size.
Let’s look at a couple of examples, where descriptive statistics may be used.

Example 1

You want to compare the number of articles that are mixed methods to those that
are not mixed methods. You decide to compute frequencies and then calculate
percentages from those frequencies.
# articles mixed = 12
# articles not mixed = 34
% mixed = 12/(12 + 34) ≈ 0.26

Example 2

You want to determine the deviation of math test scores from the class mean. You
decide to calculate the mean, standard deviation, variance, and z-score.
Scores: 44, 47, 58, 63, 74, 77, 79, 83, 84, 88, 91, 93, 96, 97
Average = (44 + 47 + 58 + 63 + 74 + 77 + 79 + 83 + 84 + 88 + 91 + 93 + 96 + 97)/14

4
 DESCRIPTIVE STATISTICS

= 1074/14 ≈ 76.71
Variance = [ (44 – 76.71)2 + (47 – 76.71)2 + … + (97 – 76.71)2 ] / (14 – 1)
   = (3976.86) / 13 ≈ 305.91
Standard deviation = square root(Variance) ≈ 17.5
z-scores for Scores:
–1.87, –1.70, –1.07, –0.78, –0.15, 0.02, 0.13, 0.36, 0.42, 0.65, 0.82, 0.93, 1.10, 1.16
So, the score of 44 is 1.87 standard deviations below the mean.
Note. Since we calculated the mean that we used for the variance calculation, the
averaging used to get the variance is reduced by 1 unit, corresponding to the information
we already accounted for by calculating the mean. This is called degrees of freedom,
corresponding to the amount of information we started with, in this case 14 scores.

WRITING A RESULTS SECTION

Scenario Used
We examined descriptive statistics for the variable of attitude to school.

Let’s now look at a sample results section write-up for reporting descriptive statistics,
using the scenario described above.
The steps we used in SPSS were:
Analyze
Descriptive Statistics
Descriptives
Options
Check Descriptives options you need (includes options of skewness and kurtosis)
Manually calculate the z-scores, which show the number of standard deviations
a score is above or below the mean. (We could have also calculated the interquartile
range, standard deviation, and variance, as we did for Example 2.)
We calculated the z-scores for the median, minimum score, and maximum score.
These were calculated as:

X − m 49 − 49.72
z= = ≈ −0.07
s 9.819

X − m 38 − 49.72
z= = ≈ −1.19
s 9.819

X − m 74 − 49.72
z= = ≈ 2.47
s 9.819

5
CHAPTER 1

The SPSS frequency output is shown below:

RESULTS SECTION WRITE-UP

The mean score for attitude to school was 49.72, with a standard deviation of
approximately 9.82 points. The minimum and maximum attitude to school scores
were 38 and 74, respectively. The mode is both 38 and 45. Since the mean is greater
than the median, and the mean is pulled towards the tail of a distribution, this
distribution is positively skewed. Refer to Figure 1.1 to see the histogram of the data.
The skewness is 0.731, with standard error of skewness of 0.123. The kurtosis
is −0.307, with standard error of kurtosis of 0.246. Skewness and kurtosis values
between ±2 indicate that the distribution may be considered to be approximately
normal. In other words, it isn’t too skewed in either direction and has an approximately
normal shape. Since the skewness and kurtosis are in the acceptable range of values,
it makes sense that the standard errors would be small. Figure 1.2 shows the overlay
of the normal curve on this distribution.
The median of 49 is approximately 0.07 standard deviations below the mean. The
minimum score of 38 is approximately 1.19 standard deviations below the mean.
The maximum score of 74 is approximately 2.47 standard deviations above the
mean. Therefore, the median is located within 68% of the scores, the minimum score
is located within 95% of the scores, and the maximum score is located within 99.7%
of the scores. In other words, the maximum score is in the interval that contains only
5% of the scores. Stated another way, this value is significantly different from the
mean of 49.72, if using 0.05 as the alpha level.

6
 DESCRIPTIVE STATISTICS

Figure 1.1. Attitude to school histogram

Figure 1.2. Attitude to school histogram with normal curve overlay

7
 CHAPTER 2

ONE-SAMPLE T-TEST

BRIEF DESCRIPTION

A one-sample t-test compares the mean of a sample to an a priori score (or

population mean). The test uses either a known population standard deviation or a
sample standard deviation. The test analyzes interval scores. Normal distribution is
not required but if data are badly skewed, a nonparametric test is preferred, such as
a binomial test with each case scored as above (1) or below (0) the a priori mean.
Note. When only a sample standard deviation is known, a t-test should be used, or if
a population standard deviation can be specified, a z-test should be used.
X −m
A t-value may be determined using the formula, t = , where s represents the
s
n
sample standard deviation. The resulting t-value may then be compared to the critical
t-value, found in the t-distribution table, for the appropriate degrees of freedom
(number of cases minus 1) and desired level of significance. If the calculated t-value
is less than the critical t-value, the null hypothesis (hypothesis of no difference)
should not be rejected, with no significant difference between the sample mean
and population mean declared. However, if the calculated t-value is larger than the
critical t-value, the null hypothesis should be rejected, with a significant difference
in means declared.
For the population standard deviation-known situation, s in the formula above is
replaced with σ, the known value, and t is replaced by z, the normal curve statistic.
Then, the normal distribution is referenced to compare the observed significance of
the z-statistic with the desired level.
A graphing calculator, excel spreadsheet, SPSS, or other software package, may
be used to calculate the t-value and determine significance. These tools may also be
used to calculate a p-value, which represents the significance of the difference. If the
p-value is less than the level of significance, which is determined by the researcher,
a significant difference may be declared (and the null hypothesis rejected). Often, a
level of 0.05 is used. Recall that .05 is the probability that we might get an unusual
sample that actually came from the distribution with the hypothesized mean m but
it was atypical and produced an extreme mean. This could happen in 5% of the
samples we drew randomly from the population.
Let’s look at a couple of examples.

9
CHAPTER 2

Example 1

You want to know if algebra exam scores of a fall freshman class are different from
a mean of 70. You will use a two-tailed test, for a 0.05 level of significance.
Scores: 42, 57, 58, 59, 59, 68, 69, 69, 72, 81, 87, 87, 93, 96, 98
Sample mean: 73
Sample standard deviation: 16.7
73 − 70
The t-value may be calculated by writing: t = , where t ≈ 0.7 .
16.7
15
The critical t-value is 2.145. (Since there are 15 scores, the number of degrees of
freedom equals 14, i.e. ( n −1 ).) Since the calculated t-value is not larger than the
critical t-value, no significant difference between the sample exam scores and score
of 70 should be declared.
A graphing calculator or software package may be used to verify the calculated
t-value, as well as to determine the p-value. The p-value is approximately 0.5. The
p-value being greater than 0.05 (the level of significance) is further evidence that the
null hypothesis should not be rejected and no significant difference declared.
In Excel, one can type into a cell the function ‘=T.DIST(0.7,14,TRUE)’, which
will return the probability .752302. Thus, .25 is the proportion of random cases that
will occur about the t-statistic. For the two-tailed test, we double this to about .5,
which gives us the probability of a difference as large as 3 points either above or
below 70. This is the cumulative probability that a t-statistic of 0.7 or smaller would
be computed from a distribution with a mean of 70 given the calculated standard
deviation and 14 degrees of freedom. For the two-tailed test at the .05 level, the
probability would have needed either to be less than .025 or greater than .975 for
significance.

Example 2

You want to compare the salaries of employees at a start-up instructional design

company to the average employee salary of employees at an established company,
which is recorded as $45,000. A 0.05 level of significance will be used.
Salaries of sample: $28,000; $30,300; $31,400; $33,600; $34,100; $40,200;
$41,800; $43,600; $44,900; $45,200; $48,300; $52,000
Mean of sample: $39,450
Standard deviation of sample: $7,771.57

39, 450 − 45, 000

t-value: t = or t ≈ −2.47
7, 771.57
12

10
 ONE-SAMPLE T-TEST

critical t-value: 2.201

p-value: p ≈ 0.03
Since the calculated t-value has an absolute value that is greater than the critical
t-value, we will reject the null hypothesis. Thus, a significant difference should
be declared. Also, the p-value of 0.03, which is less than 0.05, also shows that
a significant difference should be declared. In other words, there is a significant
difference between the salaries of employees at the startup and the average annual
salary, recorded for the established company.

WRITING A RESULTS SECTION

Scenario Used
We compared the mean of a sample of depression scores to a population mean.

Let’s now look at a sample results section write-up for analysis of the scenario
described above.
First, let’s look at the data analysis.
The steps we used in SPSS were:
Analyze
Compare Means
One-Sample T Test
Select Test Variable (depression)
Enter Test Value in Test Value box (population mean of 58)
OK

The one-sample statistics table above shows descriptive statistics for the
depression variable.

The One-Sample Test table above shows us the p-value (highlighted above) and
gives us the calculated t-value and degrees of freedom, which are needed, if a t-table
is used to determine significance, in lieu of the Sig. column.

11
CHAPTER 2

RESULTS SECTION WRITE-UP

The sample mean for depression scores was 49.86, with a standard deviation of
9.643. There is a statistically significant difference (p = 0.000) between the sample
mean and the population mean of 58. The critical t-value is 1.96 for a two-tailed test,
alpha of 0.05, and degrees of freedom of 391. The calculated t-value is −16.703.
The absolute value of the calculated t-value is greater than the critical t-value, so we
will reject the null hypothesis of no difference and declare a statistically significant
difference.

12
 CHAPTER 3

INDEPENDENT SAMPLES T-TEST

BRIEF DESCRIPTION

An independent samples t-test compares the means of two groups. The data are
interval for the groups. There is not an assumption of normal distribution (if the
distribution of one or both groups is really unusual, the t-test will not give good
results with unequal sample sizes), but there is an assumption that the two standard
deviations are equal. If the sample sizes are equal or very similar in size, even that
assumption is not critical.
X − X2
The general formula for the t-value may be written as: t = 1 .
sX − X
1 2

When the sample sizes are the same, the following formula may be used:
X1 − X 2
t= .
s12 s2 2
+
n1 n2
However, when the sample sizes are not the same, a pooled variance estimate is
X1 − X 2
used, giving the formula: t = .
2 1 1
sp  + 
 n1 n2 
A reasonable rule of thumb for deciding whether or not the standard deviations
are equal is if the ratio of the larger standard deviation to the smaller standard
deviation is less than 2. If sample sizes are equal it won’t matter, but if sample
sizes are unequal, there is a problem in interpreting the t-test if the larger standard
deviation occurs with the smaller sample size. In that case, the significance level
can be quite different from what you expect, so the researcher alpha level should be
made very small, say .001, to overcome the violation of the requirement.
Luckily, a graphing calculator or software package may be used to calculate this
t-value.
Let’s look at a couple of examples that include two independent samples and
require an independent samples t-test.

Example 1

You want to compare the statistics test scores of males and females in a freshman-
level statistics class.
13
CHAPTER 3

Male scores: 56, 58, 59, 77, 81, 88, 89, 91, 91, 92, 93, 94, 95, 96, 96
Female scores: 62, 63, 64, 67, 67, 68, 77, 81, 87, 87, 93, 93

X 1 ≈ 83.7

sx1 ≈ 14.5

n1 = 15

X 2 = 75.75

sx 2 ≈ 12

n2 = 12

Note that the standard deviations are quite similar, less than the ratio of 2 discussed
above, and the sample sizes are also similar, supporting use of the t-test.
Substituting these values into a graphing calculator, for a 2SampTTest, gives the
following summary:

t ≈ 1.5
p ≈ 0.14
df = 25

Note. For this example, the alternate hypothesis used was m1 ≠ m2 . The variance
was pooled.
Since the p-value is greater than 0.05, we may declare that there is not a significant
difference in male and female statistics test scores. The t-value of approximately
1.5 is also less than the critical t-value of 2.06, revealing the same conclusion of
failing to reject the null hypothesis. For the excel version discussed in Chapter 2,
‘=T.DIST(1.5,25,TRUE)’, the probability returned was .926931, less than the .975
we would require for two-tailed significance at .05.

Example 2

You want to compare the number of articles submitted per year by assistant professors
and associate professors at a particular university.
Number of articles submitted by assistant professors: 1, 1, 2, 2, 3, 4, 4, 4, 5, 6
Number of articles submitted by associate professors: 2, 2, 3, 4, 5, 7, 8, 8

14
 INDEPENDENT SAMPLES T-TEST

X 1 = 3.2

sx1 ≈ 1.7

n1 = 10

X 2 = 4.875

sx 2 ≈ 2.5

n2 = 8

Substituting these values into a graphing calculator, for a 2SampTTest, gives the
following summary:

t ≈ −1.7
p ≈ 0.11
df = 16

Note. For this example, the alternate hypothesis used was m1 ≠ m2 . The variance
was pooled.
Since the p-value is greater than 0.05, we may declare that there is not a significant
difference in number of articles submitted by assistant and associate professors. The
absolute value of the t-value of approximately −1.7 is also less than the critical
t-value of 2.12, revealing the same conclusion of failing to reject the null hypothesis.

WRITING A RESULTS SECTION

Scenario Used
We compared the means of males and females for the variable of somaticization.

Let’s now look at a sample results section write-up for the scenario described above.
First, let’s look at the data analysis.
The steps we used in SPSS were:
Analyze
Compare Means
Independent-Samples T Test
Enter Test Variable(s) (Somaticization)

15
CHAPTER 3

Enter Grouping Variable (sex)

Define Groups (Males and Females: used 1 and 2)
OK

The Group Statistics table above gives means of somaticization scores for males
and females, along with sample size, standard deviation, and standard error mean.

The Independent Samples Test table above shows a significant difference between
the means, so the equal variances not assumed row is used. This row shows a p-value
of .001, which indicates a significant difference between males and females for
somaticization scores.

RESULTS SECTION WRITE-UP

Levene’s test for equality of variances showed a significant difference, so we have

to look at the equal variances not assumed information (unequal variances). We have
p = 0.001, t = 3.198, and df = 375.838. So, we have that there is a statistically
significant difference, according to alpha of 0.05, between the two samples, with
males (M = 51.38) having a statistically significantly higher somaticization score
than females (M = 48.33).

16
 CHAPTER 4

PAIRED SAMPLES T-TEST

BRIEF DESCRIPTION

A paired-samples t-test compares the mean of two matched groups of people or

cases, or compares the mean of a single group, examined at two different points in
time. If the same group is tested again, on the same measure, the t-test is called a
repeated measures t-test.
As with the independent-samples t-test, the data is assumed interval for nominal
groups. The assumption of normal distribution is required for a standard t-test, although
evidence suggests nonnormality is not a problem unless skewness and kurtosis are
extreme, say over ±3 in magnitude.
Possible paired samples include husbands and wives, twins, the same group of
students in two classes, the same cadre of teachers evaluated fall and spring, the
same group of professionals at different points in their career, a group of students
taking both a pretest and a posttest, or a group of subjects taking a body mass index
assessment before and after an exercise regime.
Let’s look at a couple of examples.

Example 1

You want to compare the salaries of a group of college graduates, one year after
graduation and again five years after graduation.
One Year After: 23000, 24500, 25000, 28000, 30500, 32000, 34000, 36000,
38500, 39000, 42000, 45000
Five Years After: 24000, 25000, 26500, 29000, 31000, 33700, 36200, 38500,
40200, 40000, 43750, 48000
Using Excel, the test, “t-test: Paired Two Sample for Means” may be used to
perform the two-sample t-test.
In order to perform this test, using a TI-83, enter the salaries into lists 1 and 2,
and then highlight list 3 (L3) and enter L2 – L1. (This is done by selecting the lists
from the Edit menu.) L3 will now show all of the differences in salaries from one
year after graduation to five years after graduation. Next, choose the t-test function,
but choose “Data” for the input type. The null hypothesis ( m0 ) should be 0. The list
should show “L3.” Since the salaries are expected to increase, a one-tailed test may
be used. Thus, choose “ > m0 ,” for the type of hypothesis to test. Click “Calculate.”
Note. SPSS may also be used to compute a paired-samples t-test using the Analyze
dropdown tab with Compare Means option, Paired Samples T Test.

17
CHAPTER 4

The summary statistics are shown below:

t ≈ 6.8
p < 0.001
Since the p-value is less than 0.05 (< 0.001, in this case), we may declare there
is a significant difference in the salaries of this group, for the times of one year, and
five years, after graduation.

Example 2

You want to compare the final exam scores of a freshman class for the subjects of
U.S. History and Calculus.
U.S. History scores: 94, 67, 95, 65, 72, 60, 85, 86, 78, 76, 62, 85, 98, 72, 68, 81,
91, 92, 86, 93
Calculus scores: 100, 64, 95, 89, 75, 84, 90, 82, 79, 100, 88, 73, 83, 75, 86, 76,
83, 98, 75, 73
The summary statistics are shown below:
t ≈ 0.999
p > 0.05
Since the p-value is greater than 0.05, we may declare there is not a significant
difference in final exam scores for the two subjects, for this freshman class.

WRITING A RESULTS SECTION

Scenario Used
We compared the means of males (same group) for scores on interpersonal relations
and social stress.

Let’s now look at a sample results section write-up for the scenario described above.
First, let’s look at the data analysis.
The steps we used in SPSS were:
Analyze
Compare Means
Paired-Samples T Test
Enter paired variables (interpersonal relations and social stress)
Filter for just 1 for sex to get only males

18
 paired samples t-test

The Paired Samples Statistics table above shows descriptive statistics for the pair
of variables.

The Paired Samples Correlations table above indicates whether or not a

significant correlation exists between the variables. In this case, there is a significant
correlation, p < .01, The nonzero correlation adjusts the degrees of freedom and
standard deviation downward.

The Paired Samples Test table above shows the calculated t-value, degrees of
freedom, and p-value for the test. Since p > .05, a significant difference does not
exist for males across the variables of interpersonal relations and social stress.

RESULTS SECTION WRITE-UP

A paired samples t-test was calculated, comparing the mean scores for interpersonal
relations and social stress, for the same group of males. The paired samples
correlations table shows that a significant correlation exists between the variables
of interpersonal relations and social stress. In other words, how someone scores on
social stress is likely to be similar to how they will score on interpersonal relations.
The paired differences table shows a t-value of 0.326, df = 228, and p-value of .745.
Since p > .05, we can declare that there is not a statistically significant difference
between the means for the males for scores on interpersonal relations and social
stress.

19
 CHAPTER 5

ONE-WAY ANOVA

BRIEF DESCRIPTION

A one-way ANOVA (analysis of variance) compares the means of two or more groups
for one dependent variable. A one-way ANOVA is required when the study includes
more than two groups. (In other words, a t-test cannot be used.) As with t-tests,
there is one independent variable and one dependent variable. Interval dependent
variables for nominal groups are required. The assumption of normal distribution is
not required.
When performing a one-way ANOVA, a minimum sample size of 30 is desired,
but equal numbers per group is not required. ANOVA compares the variation within
a group (on average) to the equivalent variation based on group means’ variation.
Typically, the ratio of between-group variation to within-group variation should be
greater than 3 or 4, although smaller values will be significant for large sample sizes.
Having a sample size greater than or equal to 30 reduces the risk of making a Type
II error (failing to reject the null hypothesis, when it should be rejected, or failing
to declare a statistically significant difference, when you should have.) In order to
reduce the risk of making a Type I error, the researcher may choose to lower the
alpha level, or level of significance.
Let’s look at a couple of examples.

Example 1

You want to compare the final math averages of students who are taught using a
teacher-directed, student-centered, or combination approach.
The Excel spreadsheet below shows the data:

21
Another random document with
no related content on Scribd:
929. Orelli, Inscpt. Latinar. selectar. Turin, 1828, vol. I. pp. 406-412.

930. See Cumont, T. et M. II. p. 95, inscr. 15, p. 98, inscr. 23; p. 100
inscr. 40; p. 101, inscr. 41. The tomb of Vincentius in the
Catacomb of Praetextatus at Rome would show an instance
of the joint worship of Sabazius, the consort of the Great
Mother, and of Mithras, if we could trust Garrucci’s restoration,
for which see his Les Mystères du Syncrétisme Phrygien,
Paris, 1854. It has been quoted in this sense by Hatch, H.L. p.
290; but Cumont, T. et M. II. pp. 173 and 413, argues against
this construction. For the pictures themselves, see Maass,
Orpheus, München, 1895, pp. 221, 222.

931. Cumont, T. et M. II. p. 261, Fig. 99.

932. Kenyon, Gk. Papyri, p. 65.

933. This is the more likely because his second initiator bears the
name of Asinius, which, as he himself says (Apuleius,
Metamorph. Bk XI. c. 27), was not unconnected with his own
transformation into the shape of an ass. The Emperor
Commodus was initiated into both religions (Lampricius,
Commodus, c. IΧ.).

934. See n. 1, p. 259, supra.

935. Dill, Nero to Marcus Aurelius, p. 625, n. 3, quoting Gasquet,

Mithras, p. 137. See also Gibbon, vol. III. p. 498, Bury
(Appendix 15).

936. Justin Martyr, First Apology, c. LXVI.

937. Porphyry, de antro nymph. c. 15. Tertullian, de Praescpt. c.

40.

938. Porphyry, op. et loc. cit.

939. See Cumont, T. et M. I. p. 339, for authorities.

940. Augustine, In Johann. evang. tractatus, VII. or Cumont, T. et
M. II. p. 59. This last thinks it more probable that the passage
refers to Attis, as there is an allusion in it to redemption by
blood. But this would hardly apply to the self-mutilation of the
Galli, while it would to the blood-bath of the Taurobolium and
Criobolium which so many high initiates of Mithras boast of
undergoing.

941. J. Maurice, “La Dynastie Solaire des Seconds Flaviens,” Rev.

Archeol. t. XVII. (1911), p. 397 and n. 1.

942. Cumont, T. et M. I. p. 339, quoting Minucius Felix.

943. Op. cit. I. p. 65.

944. The remains of five Mithraea were found in Ostia alone.

945. Cumont, T. et M. II. p. 204, Fig. 30, and p. 493, Fig. 430; or
P.S.B.A. 1912, Pl. XIII. Figs. 1 and 2.

946. Cumont, T. et M. I. p. 62.

947. The story quoted from Pseudo-Augustine (Cumont, op. cit. I.

p. 322) about the hands of the initiates being bound with
chickens’-guts which were afterwards severed by a sword
might account for the number of birds’ bones.

948. Cumont, op. cit. II. p. 21, gives the passage from Lampridius
mentioned in n. 1, p. 260, supra.

949. Op. cit. I. p. 322, quoting Zacharius rhetor.

950. See Chapter II, Vol. I. p. 62, supra.

951. Cumont, T. et M. II. p. 18, for the passage in St Jerome in

which these degrees are enumerated. They all appear in the
inscriptions given by Cumont, except that of Miles or Soldier.
An inscription by two “soldiers” of Mithras has, however, lately
been found at Patras and published by its discoverers, M.
 Charles Avezou and M. Charles Picard. See R.H.R. t. LXIV.
(1911), pp. 179-183.

952. Cumont, T. et M. I. pp. 315 sqq.

953. Tertullian, de Corona, c. 15.

954. Porphyry, de antro nymph. c. 15.

955. Cumont, T. et M. I. p. 322. Gregory of Nazianza (A.D. 320-390)

is the first authority for these tortures (κολάσεις) in point of
time. Nonnus the Mythographer gives more details, but is
three centuries later.

956. Renan, Marc-Aurèle, p. 577.

957. Cumont, T. et M. I. p. 73.

958. Op. cit. II. p. 294, Fig. 149; p. 298, Fig. 154; p. 300, Fig. 156;
p. 304, Fig. 161; p. 488, Fig. 421.

959. Op. cit. I. p. 175, Fig. 10.

960. Op. cit. I. p. 39, n. 6, quoting the Arda Viraf namak. A

quotation from Arnobius, adv. gentes, which follows, merely
says that the Magi boast of their ability to smooth the
believers’ passage to heaven.

961. See Chap. VIII, p. 74, n. 3, supra.

962. That those who had taken the degree of Pater were called
ἀετοί or eagles appears from Porphyry, de Abstinentia, Bk IV.
c. 16. Cumont doubts this; see T. et M. I. p. 314, n. 8. The idea
probably had its origin in the belief common to classical
antiquity that the eagle alone could fly to the sun, from which
the Mithraist thought that the souls of men came, and to which
those of perfect initiates would return. Cf. op. cit. I. p. 291.

963. Lafaye, L’Initiation Mithriaque, p. 106.

964. Cumont, T. et M. II. p. 56.

965. Porphyry, de Abstinentia, Bk IV. c. 16 says this was so.

966. Cumont, T. et M. I. p. 318, n. 1, points out that an initiate might

become Pater Patrum immediately after being made Pater or
Pater sacrorum simply. This appears from the two monuments
both dated the same year of Vettius Agorius Praetextatus, op.
cit. II. p. 95.

967. See Ammianus Marcellinus Bk XXII. c. 7, for his life under

Julian. His career is well described by Dill, Roman Society in
the Last Century of the Western Empire, 1899, pp. 17, 18, 30,
154, 155.

968. Cumont, T. et M. II. p. 100, inscr. 35; p. 98, inscr. 24.

969. Op. cit. II. p. 130, inscr. 225; p. 132, inscr. 239; p. 134, inscr.
257. The two decurions may of course have been decurions
of the rite only, as to which see op. cit. I. p. 326.

970. Op. cit. I. p. 324: Tertullian, Praescpt. c. 40.

971. Cumont, T. et M. I. p. 65. Thirty-five seems to be the greatest

number belonging to any one chapel.

972. Op. cit. I. p. 327.

973. Amm. Marcell. passim.

974. Neander, Ch. Hist. III. p. 136.

975. Marinus, vita Procli, pp. 67, 68; Neander, op. cit. III. p. 136.

976. Witness the reduction of Mitra, who plays such an important

part in the religion of the Vedas, to the far lower position of
chief of the Izeds or Yazatas in the Sassanian reform.

977. Cumont, T. et M. I. p. 250, for authorities.

978. Gibbon, Decline and Fall (Bury’s ed.), I. p. 260 n. 106.

979. Reville, Religion sous les Sevères, p. 102.

980. Cumont, T. et M. I. p. 347.

981. Dill, Last Century, etc. p. 29, n. 2.

982. Cumont, T. et M. I. p. 347.

983. Op. cit. I. pp. 329, 330; Dill, Nero to Marcus Aurelius, p. 624.

984. Cumont, “L’aigle funéraire des Syriens et l’apothéose des

empereurs.” R.H.R., 1910, pt ii. pp. 159 sqq.
985. Cf. the “solitary eagle” of the Magic Papyrus quoted on p. 265
supra.

986. Maury, La Magie et L’Astrologie, passim. The Zend Avesta

also denounces magic as did the later Manichaeism. See p.
342 infra.

987. As in Shakespeare’s Macbeth.

988. So Cumont, T. et M. I. pp. 45, 349, 350. He seems to rely,

however, entirely on the passage in the Acta Archelai (as to
which see n. 1, p. 280 infra), wherein the supposed bishop
Archelaus addresses the equally imaginary Manes as
“Savage priest and accomplice of Mithras!”—possibly a mere
term of abuse. See Hegemonius, Acta Archelai, ed. Beeson,
Leipzig, 1906, c. XL. p. 59.

989. Cumont, T. et M. I. p. 41. He sees in the scenes which border

the Tauroctony references or parallels to the fig-leaves of
Genesis, the striking of the rock by Moses, and the ascension
of Elijah. In the so-called Mithraic Ritual of the Magic Papyrus
of Paris, there are certain Hebrew words introduced, such as
πιπι (a well-known perversion of the Tetragrammaton),
σανχερωβ and σεμες ιλαμ (The “Eternal Sun”).

990. See the story which Josephus, Antiq. XX. cc. 2, 3, 4, tells
about Izates, king of Adiabene, who wanted to turn Jew and
thereby so offended his people that they called in against him
Vologeses or Valkash, the first reforming Zoroastrian king and
collector of the books of the Zend Avesta. Cf. Darmesteter,
The Zend Avesta (Sacred Books of the East), Oxford, 1895,
p. xl. Cf. Ém. de Stoop La Diffusion du Manichéisme dans
l’Empire romain, Gand, 1909, p. 10.

991. Circa 296, A.D. See Neander, Ch. Hist. II. p. 195, where the
authenticity of the decree is defended. For the provocation
given to the Empire by the anti-militarism of Manes see de
Stoop, op. cit. pp. 36, 37.
992. Al-Bîrûnî, Chronology of Ancient Nations, p. 190. The date he
gives is twelve years before the accession of Ardeshîr. E.
Rochat, Essai sur Mani et sa Doctrine, Genève, 1897, p. 81,
examines all the different accounts and makes the date from
214 to 218 A.D.

993. Epiphanius, Haer. LXVI. c. 1, p. 399, Oehler; Socrates, Hist.

Eccl. Bk I. c. 22; Hegemonius, Acta Archelai, c. LXIV.

994. Muhammed ben Ishak, commonly called En-Nadîm, in the

book known as the Fihrist, translated by Flügel, Mani, seine
Lehre und seine Schriften, Leipzig, 1862, pp. 83, 116, 118,
119. Cf. Rochat, op. cit. p. 75.

995. Al-Bîrûnî, Chronology, p. 190.

996. Flügel, op. cit. p. 84; Rochat, op. cit. p. 83.

997. Hegemonius, Acta Arch. c. XL., p. 59, Beeson. Rochat, op. cit.
pp. 9-49, discusses the authenticity of the Acta chapter by
chapter. He thinks the pretended discussion between
Archelaus and Manes unhistorical, and the account of it
possibly modelled on that between St Augustine and Faustus
the Manichaean. The remainder of the Acta he considers
fairly trustworthy as an account of Manes’ own tenets. This
may well be, as Epiphanius, Haer. LXVI. cc. 6-7, 25-31,
transcribes the epistle to Marcellus, its answer, and the
exposition of Turbo, and could scarcely have heard, as early
as 375 A.D., about which time he wrote, of St Augustine’s
discussion. The Acta owe much to the care of the American
scholar, Mr Beeson of Chicago, who has given us the careful
edition of them mentioned in n. 1, p. 277 supra. It is a pity that
he did not see his way to keep the old numeration of the
chapters.

998. Beausobre, Hist. du Manichéisme, Paris, 1734, Pt I. Bk II. cc.

1-4. Cf. Stokes in Dict. Christian Biog. s.v. Manes; Rochat, op.
cit. p. 83.
999. Rochat, op. cit. p. 89.

1000.
Abulfarag in Kessler, Forschungen über die Manichäische
Religion, Berlin, 1889, Bd I. p. 335; Rochat, op. cit. p. 84;
Neander, Ch. Hist. II. p. 168.

1001.
Flügel, op. cit. p. 85. Cf. Al-Bîrûnî, India (ed. Sachau), p. 55,
where Manes quotes the opinion of Bardesanes’ “partizans.”
There are many words put into the mouth of Manes in the
work quoted which argue acquaintance with the Pistis Sophia.

1002.
Abulmaali in Kessler, op. cit. p. 371; Firdaûsi, ibid. p. 375;
Mirkhônd, ibid. p. 379. Cf. Rochat, op. cit. p. 81. He is said to
have painted his pictures in a cave in Turkestan (Stokes in
Dict. Christian Biog. s.v. Manes), which would agree well
enough with the late German discoveries at Turfan, for which
see A. von Le Coq in J.R.A.S. 1909, pp. 299 sqq.

1003.
Flügel, op. cit. p. 85.

1004.
Al-Jakûbi in Kessler, op. cit. pp. 328, 329; cf. Rochat, op. cit.
p. 88.

1005.
Al-Bîrûnî, Chronology, pp. 191, 192.

1006.
Rochat, op. cit. p. 89. Al-Bîrûnî, whom he quotes, however,
says merely that the Manichaeans increased under Ormuz,
and also that Ormuz “killed a number of them.” See last note.
1007.
Al-Jakûbi in Kessler, op. cit. p. 330. But Darmesteter (see
passage quoted in n. 2, p. 284 infra) puts this event as
happening after Ormuz’ death and under Shapur II.

1008.
Al-Bîrûnî, Chronology, p. 191. The town is called Djundi-sâbur
or Gundisabur.

1009.
Al-Jakûbi, ubi cit. supra; Eutychius quoted by Stokes, Dict.
Christian Biog. s.v. Manes.

1010.
Rochat, op. cit. p. 93, examines all the evidence for this and
comes to the conclusion given in the text.

1011. Malcolm, History of Persia, London, 1821, Vol. I. pp. 95, 96.

1012.
G. Rawlinson, The 6th Oriental Monarchy, 1873, p. 222;
Rochat, op. cit. p. 53.

1013.
See Chap. XII supra, p. 232.

1014.
See n. 1, p. 278 supra.

1015.
Al-Bîrûnî, Chron. p. 187, makes Manes the successor or
continuator of Bardesanes and Marcion. This was certainly
not so; but it was probably only from their followers that he
derived any acquaintance with Christianity. See n. 7, p. 280
supra. So Muhammad or Mahommed, four centuries later,
drew his ideas of the same faith from the heretics of his day.
1016.
Rhys Davids, Buddhist India, 1903, p. 318, says that after 300
A.D. Buddhism was everywhere in decay in India.

1017.
Rochat, op. cit. p. 58.

1018.
Darmesteter, Zend Avesta, pp. xl, xli.

1019.
Op. cit. pp. xlvii sqq.

1020.
Al-Bîrûnî, Chron. p. 192.

1021.
Elisaeus Vartabed in Langlois’ Collection des Hist. de
l’Arménie, Paris, 1868, t. II. p. 190. The story is repeated
almost word for word by Eznig of Goghp, ibid. p. 875. Cf.
Neander, Ch. Hist. II. p. 171.

1022.
Rochat, op. cit., following Kessler, shows, it seems,
conclusively, that this is another name for Manes’ father,
Fatak or Patecius.

1023.
She was a courtezan at Hypselis in the Thebaid according to
Epiphanius, Haer. LXVI. c. 11, p. 400, Oehler. As Baur, Die
Manichäische Religionssystem, Tübingen, 1831, p. 468 sqq.
has pointed out, this is probably an imitation of the story told
about Simon Magus and his Helena (see Chap. VI supra). It
seems to have arisen as an embroidery, quite in Epiphanius’
manner, upon the story in the Acta, that Scythianus married a
captive from the Upper Thebaid (Hegemonius, op. cit. c. LXII.
p. 90, Beeson).
1024.
Many guesses have been made as to the allusions concealed
under these names, as to which see Rochat, op. cit. pp. 64-
73. Neander (Ch. Hist. II. p. 16) quotes from Ritter the
suggestion that Terebinthus may come from an epithet of
Buddha, Tere-hintu “Lord of the Hindus.” One wonders
whether it might not have been as fitly given to a Jewish slave
sold at the Fair of the Terebinth with which Hadrian closed his
war of extermination.

1025.
These four books may have been intended for the
Shapurakhan, the Treasure, the Gospel and the Capitularies,
which Al-Bîrûnî, Chron. p. 171, attributes to Mani. Cf.
Epiphanius, Haer. LXVI. c. 2, p. 402, Oehler, and the Scholia of
Théodore bar Khôni in Pognon, Inscriptions Mandaïtes des
Coupes de Khouabir, pp. 182, 183.

1026.
Epiphanius, op. cit. c. 1, p. 398, Oehler.

1027.
Colditz in Kessler, op. cit. pp. 15, 16. Cf. Rochat, op. cit. pp.
65, 66.

1028.
Morrison, Jews under Romans, p. 325 for authorities. Philo,
de Vit. Contempl. etc. c. III. says that similar communities
existed in his time near the Mareotic lake in Egypt. But the
date of the treatise and its attribution to Philo are alike
uncertain. The first mention of Buddha in Greek literature is
said to be that by Clem. Alex. Strom. Bk I. c. 15.

1029.
Harnack in Encyc. Britann. 9th edition, s.v. Manichaeans, p.
48, says “There is not a single point in Manichaeism which
demands for its explanation an appeal to Buddhism.” This
may be, but the discoveries at Turfan and Tun-huang have
 made a connection between the two more probable than
appeared at the time he wrote. See also Kessler as quoted by
Rochat, op. cit. pp. 192, 193.

1030.
This appears from the Chinese Treatise at Pekin mentioned
later. See p. 293, n. 2.

1031.
Rochat, op. cit. p. 194. So Socrates, Eccl. Hist. Bk I. c. 22,
calls Manichaeism “a sort of heathen (Ἑλληνίζων)
Christianity.”

1032.
Hegemonius, Acta, c. VII. p. 91, Beeson; Flügel, op. cit. p. 86.

1033.
Certainly none is recorded in the Christian accounts, where
Darkness is called Hyle or Matter. En Nadîm (Flügel, op. cit.
p. 86) makes Manes call the good God “the King of the
Paradise of Light” and (p. 90) the Spirit of Darkness,
Hummâma. Schahrastâni, as quoted in Flügel’s note (p. 240),
makes this word mean “mirk” or “smoke” (Qualm). It would be
curious if Hummâma had any connection with the Elamite
Khumbaba, the opponent of the Babylonian hero Gilgamesh,
because this personage already figures in Ctesias’ story
about Nannaros, which has been recognized as a myth
relating to the Moon-god.

1034.
τὸ τῆς ὕλης δημιούργημα Hegemonius, Acta, c. VIII. p. 9,
Beeson. Cf. Alexander of Lycopolis, adv. Manichaeos, c. II.

1035.
Epiph. Haer. LXVI. c. 6, p. 408, Oehler; Hegemonius, Acta, c. V.
pp. 5-7, Beeson. The authenticity of the letter is defended by
Kessler, op. cit. p. 166. Cf. Rochat, op. cit. p. 94 contra.
1036.
τῶν κακῶν ἐπὶ τὸν θεὸν ἀναφέρουσιν, ὧν τὸ τέλος κατάρας
ἐγγύς. It is evidently intended for a quotation from Heb. vi. 8,
which however puts it rather differently as ἐκφέρουσα δὲ
ἀκάνθας καὶ τριβόλους ἀδόκιμος καὶ κατάρας ἐγγύς, ἧς τὸ
τέλος εἰς καῦσιν. “But that which beareth thorns and briers is
to be rejected and is nigh unto cursing; whose end is to be
burned.” The Khuastuanîft or Manichaean confession
mentioned later repeats this phrase about God not being the
creator of evil as well as of good. See p. 335 infra.

1037.
Hegemonius, Acta, c. VII. p. 9, Beeson.

1038.
En Nadîm in Kessler, op. cit. pp. 386, sqq. Kessler’s
translation of En Nadîm, which is given in the first Appendix to
the work quoted, differs slightly from that of Flügel and
depends on a somewhat better text than the last-named. It is
therefore used when possible in the remaining notes to this
chapter. Flügel’s book, however, has the advantage of a
commentary of some 300 pages marked with great erudition,
and must still be consulted by anyone wishing to be
acquainted with its subject.

1039.
Plutarch, de Is. et Os. c. XLV., says, however, that “evil must
have a principle of its own,” so that it cannot be the work of a
benevolent being. As he is generally supposed to have taken
his account of the Persian teaching from Theopompos of
Chios, who was at the Court of Ptolemy about 305 B.C., his
evidence is against those who, like M. Cumont, would make
the “Zervanist” opinion, which assumes a common principle
for good and evil, pre-Christian. Yet the point does not yet
seem capable of decision, as Plutarch may here be only
giving us his own opinion.
1040.
Casartelli, op. cit. p. 44.

1041.
This is really the crux of the whole question. If the idea could
be traced back to the philosophers of Ionia (e.g. Heraclitus of
Ephesus) and their theory of eternal strife and discord being
the cause of all mundane phenomena, it is difficult to say
whence the Ionians themselves derived it, save from Persia.
We can, of course, suppose, if we please, that the Persians
did not invent it de novo, but took it over from some of their
subjects. Among these, the Babylonians, for instance, from
the earliest times portrayed their demons as not only
attempting to invade the heaven of the gods, but as being in
perpetual warfare with one another. But the very little we know
of Babylonian philosophy would lead us to think that it inclined
towards pantheism of a materialistic kind rather than to
dualism.

1042.
En Nadîm, in Kessler, op. cit. p. 387; Flügel, op. cit. p. 86.

1043.
The likeness of this to the cosmogony of the Ophites and their
successor Valentinus is of course marked (cf. Chaps. VIII and
IX supra). Manes may have borrowed it directly from
Valentinus’ follower Bardesanes, whose doctrines were
powerful in Edessa and Mesopotamia in his time, or he may
have taken it at first-hand from Persian or Babylonian
tradition. That Manes was acquainted with Bardesanes’
doctrines, see n. 7, p. 280 supra.

1044.
En Nadîm in Kessler, op. cit. p. 387; Flügel, op. cit. p. 86.
Flügel’s text adds to these members other “souls” which he
names Love, Belief, Faith, Generosity, and Wisdom. Kessler
substitutes Courage for Generosity and seems to make these
“souls” the members’ derivatives.
1045.
See last note.

1046.
See Chapter XII, p. 251 supra. Here, again, the traditional and
monstrous figure of Satan may have been copied from the
sculptured representations of the composite demons of
Babylonia (e.g. Rogers, Religion of Babylonia and Assyria,
Frontispiece and Figs. 1 and 13). Yet if we take the Mithraic
lion, as M. Cumont would have us do, as the symbol of fire
and the serpent as that of the earth, we have in the five sorts
of animals the five στοιχεῖα or elements of Aristotle. Cf.
Aetius, de Placitis Philosophorum, ed. Didot, Bk I. c. iii. § 38
(Plutarch, Moralia, II.), p. 1069. Yet the nearest source from
which Manes could have borrowed the idea is certainly
Bardesanes, who, according to Bar Khôni and another Syriac
author, taught that the world was made from five substances,
i.e. fire, air, water, light and darkness. See Pognon, op. cit. p.
178; Cumont, La Cosmogonie Manichéenne d’après
Théodore bar Khôni, Bruxelles, 1908, p. 13, n. 2.

1047.
En Nadîm in Kessler, op. cit. p. 388; Flügel, op. cit. p. 87. As
the ancients were unacquainted with the properties of gases,
it is singular that they should have formed such a conception
as that of the compressibility and expansibility of spirits. Yet
the idea is a very old one, and the Arabian Nights story of the
Genius imprisoned in a brass bottle has its parallel in the
bowls with magical inscriptions left by the Jews on the site of
Babylon (Layard, Nineveh and Babylon, 1853, pp. 509 sqq.),
between pairs of which demons were thought to be
imprisoned. Cf. Pognon, op. cit. p. 3. Something of the kind
seems indicated in the “Little Point,” from which all material
powers spring, referred to by Hippolytus and the Bruce
Papyrus.
1048.
So in the Pistis Sophia, it is the “last Parastates” or assistant
world who breathes light into the Kerasmos, and thus sets on
foot the scheme of redemption. Cf. Chapter X, p. 146 supra.

1049.
Yet the Fundamental Epistle speaks of the twelve “members”
of God, which seem to convey the same idea See Aug. c. Ep.
Fund. c. 13.

1050.
Thus En Nadîm in Kessler, op. cit. pp. 388, 389; Flügel, op.
cit. p. 87. But here the Christian tradition gives more details
than the Mahommedan. Hegemonius, Acta, c. VII., p. 10,
Beeson, and Bar Khôni (Pognon, p. 185), are in accord that
the God of Light produced from himself a new Power called
the Μήτηρ τῆς Ζωῆς or Mother of Life, that this Mother of Life
projected the First Man, and that the First Man produced the
five elements called also his “sons,” to wit, wind, light, water,
fire and air, with which he clothed himself as with armour. See
Cumont, Cosmog. Manich. p. 16, n. 4, for the harmonizing of
the texts [N.B. the omission of πῦρ from his quotation from
the Acta is doubtless a clerical error]. The identification of the
Mother of Life with the “Spirit of the Right [Hand]” is accepted
by Bousset, Hauptprobleme, pp. 177, 178, and may be
accounted for by the crude figure by which the Egyptians
explained the coming-forth of the universe from a single male
power. See Budge, Hieratic Papyri in the Brit. Mus. p. 17.

1051.
These were also the “sons” of Darkness or Satan. See Bar
Khôni (Pognon, p. 186). The reason that led the God of Light
to send a champion into the lists was, according to Bar Khôni
(Pognon, p. 185), that the five worlds of his creation were
made for peace and tranquillity and could therefore not help
him directly in the matter. Cf. St Augustine, de Natura Boni, c.
XLII. But Manes doubtless found it necessary to work into his
 system the figure of the First Man which we have already
seen prominent in the Ophite system. Cumont, Cosmog.
Manich. p. 16, says few conceptions were more widely spread
throughout the East. It is fully examined by Bousset,
Hauptprobleme, in his IVth chapter, “Der Urmensch.” The First
Man is, in the Chinese treatise lately found at Tun-huang in
circumstances to be presently mentioned, identified with the
Persian Ormuzd and the five elements are there declared to
be his sons. See Chavannes and Pelliot, Un Traité Manichéen
retrouvé en Chine, pt 1, Journal Asiatique, série X., t. XVIII.
(1911), pp. 512, 513. The 12 elements which helped in his
formation seem to be mentioned by no other author than En
Nadîm. St Augustine, however, Contra Epistulam Fundamenti,
c. 13, speaks of the “12 members of light.” The Tun-huang
treatise also mentions “the 12 great kings of victorious form”
whom it seems to liken to the 12 hours of the day. As the
Pistis Sophia does the same with the “12 Aeons” who are
apparently the signs of the Zodiac, it is possible that we here
have a sort of super-celestial Zodiac belonging to the
Paradise of Light, of which that in our sky is a copy. It should
be remembered that in the Asiatic cosmogonies the fixed
stars belong to the realm of good as the representatives of
order, while the planets or “wanderers” are generally evil.

1052.
En Nadîm in Kessler, op. cit. p. 389; Flügel, op. cit. pp. 87, 88.
According to the Christian tradition, the Powers of Darkness
devoured only the soul of the First Man which was left below
when his body, as will presently be seen, returned to the
upper world. See Hegemonius, Acta, c. VII., p. 10, Beeson.

1053.
Both the Christian and the Mahommedan traditions agree as
to this result of the fight, which is paralleled not only by the
more or leas successful attempt of Jaldabaoth and his powers
to eat the light of Pistis Sophia, but also by a similar case in
orthodox Zoroastrianism. For all these see Cumont, Cosmog.
 Manich. p. 18, n. 4. Bar Khôni (Pognon, p. 186), goes further
and describes the surrender of the First Man as a tactical
effort on his part, “as a man who having an enemy puts
poison in a cake and gives it to him.” Alexander of Lycopolis
(adv. Manich. c. III.), on the other hand declares that God
could not avenge himself upon matter (as he calls Darkness)
as he wished, because he had no evil at hand to help him,
“since evil does not exist in the house and abode of God”; that
he therefore sent the soul into matter which will eventually
permeate it and be the death of it; but that in the meantime
the soul is changed for the worse and participates in the evil
of matter, “as in a dirty vessel the contents suffer change.”
These, however, are more likely to be the ideas of the
Christian accusers than the defences of the Manichaean
teachers.

1054.
En Nadîm in Kessler, op. cit. pp. 389, 390; Flügel, op. cit. p.
87. As Kessler points out, En Nadîm gives two accounts
doubtless taken from different Manichaean sources. In one,
he says simply that the King of the Paradise of Light followed
with other gods and delivered the First Man, the actual victor
over Darkness being called “the Friend” of the Lights (like
Mithras). He then goes on to say that Joy (i.e. the Mother of
Life) and the Spirit of Life went to the frontier, looked into the
abyss of hell and saw the First Man and his powers were held
enlaced by Satan, “the Presumptuous Oppressor and the Life
of Darkness”; then she called him in a loud and clear voice,
and he became a god, after which he returned and “cut the
roots of the Dark Powers.” For Bar Khôni’s amplification of
this story see p. 302, n. 1, and p. 324 infra. The whole of this,
together with the cutting of the roots, is strongly reminiscent of
the Pistis Sophia.

1055.
En Nadîm in Kessler, op. cit. pp. 391, 392; Flügel, op. cit. p.
98. The Acta (Hegemonius, op. cit. c. VIII., p. 11, Beeson) say
that the “Living Spirit” before mentioned “created the Cosmos,
descended clothed with three other powers, drew forth the
rulers (οἱ ἄρχοντες) and crucified them in the firmament which
is their body the Sphere.” “Then he created the lights
(φωστῆρες) which are the remnants of the soul, caused the
firmament to encompass them, and again created the earth
[not the Cosmos] with its eight aspects.” The Latin version
after “earth” adds “they (sic!) are eight.” which if it refers to the
aspects would agree with En Nadîm. Alexander of Lycopolis
(adv. Manich. c. III.), who had been a follower of Manes and
was a Christian bishop some 25 years after Manes’ death,
says that “God sent forth another power which we call the
Demiurge or creator of all things; that this Demiurge in
creating the Cosmos separated from matter as much power
as was unstained, and from it made the Sun and Moon; and
that the slightly stained matter became the stars and the
expanse of heaven.” “The matter from which the Sun and
Moon were taken,” he goes on to say, “was cast out of the
Cosmos and resembles night” [Qy the Outer Darkness?],
while the rest of the “elements” consists of light and matter
unequally mingled. Bar Khôni (Pognon, op. cit. p. 188), as will
presently be seen, says that the Living Spirit with the Mother
of Life and two other powers called the Appellant and
Respondent [evidently the “three other powers” of the Acta]
descended to earth, caused the Rulers or Princes to be killed
and flayed, and that out of their skins the Mother of Life made
11 heavens, while their bodies were cast on to the earth of
darkness and made 8 earths. The Living Spirit then made the
Sun, the Moon, and “thousands of Lights” (i.e. Stars) out of
the light he took from the Rulers. That this last story is an
elaboration of the earlier ones seems likely, and the flaying of
the Rulers seems to be reminiscent of the Babylonian legend
of Bel and Tiamat, an echo of which is also to be found in the
later Avestic literature. See West, Pahlavi Texts (S.B.E.), pt iii.
p. 243. Cf. Cumont, Cosmog. Manich. p. 27, n. 2.
1056.
En Nadîm in Kessler, op. cit. p. 392; Flügel, op. cit. pp. 89-90.
This would agree perfectly with the system of the Pistis
Sophia, where it is said that the “receivers of the Sun and
Moon” give the particles of the light as it is won from matter to
Melchizedek, the purifier, who purifies it before taking it into
the Treasure-house (pp. 36, 37, Copt.). The idea that the
Sun’s rays had a purifying effect shows shrewd observation of
nature before his bactericidal power was discovered by
science. So does the association of the Moon with water,
which doubtless came from the phenomenon of the tides. Is
the Column of Glory the Milky Way?

1057.
The Ecpyrosis or final conflagration is always present in
orthodox Mazdeism, where it inspires its Apocalypses, and is
in effect the necessary conclusion to the drama which begins
with the assault on the world of light by Ahriman. For
references, see Söderblom, op. cit. chap. IV. From the
Persians it probably passed to the Stoics and thus reached
the Western world slightly in advance of Christianity. “The day
when the Great Dragon shall be judged” is continually on the
lips of the authors of the Pistis Sophia and the Μέρος τευχῶν
Σωτῆρος, and the conception may therefore have reached
Manes from two sources at once. The angels maintaining the
world as mentioned in the text are of course the
Splenditenens and Omophorus about to be described.

1058.
Hegemonius, Acta, c. VIII. p. 12, Beeson. St Augustine (contra
Faustum, Bk XX. c. 10) mentions the Wheel briefly and rather
obscurely. It seems to have fallen out of the account of Bar
Khôni. But see the Tun-huang treatise (Chavannes et Pelliot,
op. cit. 1ère partie, pp. 515, n. 2, 516, 517, n. 3). There can be
little doubt that it is to be referred to the Zodiac. The Aeons of
the Light seem to be the five worlds who here play the part of
the Parastatae in the Pistis Sophia.