CHAPTER 3
METHODOLOGY OF THE JAPA EXPERIMENT
Introduction
Below
are the hypotheses for this experiment. Since these hypotheses were
statistically evaluated, they should, strictly speaking, be stated in terms of
null hypotheses, because statistical techniques actually test null hypotheses by
analyzing the strength of the sample evidence against the null hypotheses. That
is, statistical methods assess whether a null hypothesis can be shown to be
inconsistent with the observed data, and the null hypothesis is rejected or not
rejected based on statistical tests. The practical purpose of this research,
however, is to gauge the amount of support that the data provide for directional
research hypotheses. Therefore, the hypotheses are stated in such terms. For
instance, the first hypothesis of this study is that, at the .05 level the maha
mantra group will show significantly decreased stress from pretest to
posttest compared with the alternate mantra and control groups. If this
hypothesis were stated in terms of a null hypothesis, it would state that there
will be no difference between the change in stress scores from pretest to
posttest between the maha mantra group and the other groups. Agresti and
Finlay (1986) confirm that it is acceptable to state research hypotheses in
terms of alternate hypotheses, instead of in terms of null hypotheses, when
performing statistical tests.
The
primary hypotheses of this study were that, at alpha = .05:
1)
The intervention group will show significantly decreased stress from pretest to
posttest compared with both the alternate mantra and control groups.
2)
The intervention group will show significantly decreased depression from
pretest to posttest compared with both the alternate mantra and control groups.
3)
The intervention group will show significantly increased sattva from
pretest to posttest compared with both the alternate mantra and control groups.
4)
The intervention group will show significantly decreased rajas from
pretest to posttest compared with both the alternate mantra and control groups.
5)
The intervention group will show significantly decreased tamas from
pretest to posttest compared with both the alternate mantra and control groups.
Secondary
hypotheses of this study were that, at alpha = .05:
6)
The intervention group will show significantly decreased stress from
pretest to follow-up compared with the alternate mantra and control groups,
though some decrement in the amount of decrease for the maha mantra
group, compared with the other two groups, is expected in pretest-follow-up
scores compared with pretest-posttest scores.
7)
The intervention group will show significantly decreased depression from
pretest to follow-up compared with the alternate mantra and control groups,
though some decrement in the amount of decrease for the maha mantra group,
compared with the other two groups, is expected in pretest-follow-up scores
compared with pretest-posttest scores.
8)
The intervention group will show significantly increased sattva from
pretest to follow-up compared with the alternate mantra and control groups,
though some decrement in the amount of increase for the maha mantra group,
compared with the other two groups, is expected in pretest-follow-up scores
compared with pretest-posttest scores.
9)
The intervention group will show significantly decreased rajas from
pretest to follow-up compared with the alternate mantra and control groups,
though some decrement in the amount of decrease for the maha mantra group,
compared with the other two groups, is expected in pretest-follow-up scores
compared with pretest-posttest scores.
10)
The intervention group will show significantly decreased tamas from
pretest to follow-up compared with the alternate mantra and control groups,
though some decrement in the amount of decrease for the maha mantra group,
compared with the other two groups, is expected in pretest-follow-up scores
compared with pretest-posttest scores.
11)
Gender will have no effect on stress, depression, sattva, rajas,
or tamas, from pretest to posttest or from pretest to follow-up.
12)
Age will have no effect on stress, depression, sattva, rajas, or
tamas, from pretest to posttest or from pretest to follow-up..
13)
Chanting frequency for the maha mantra group will be positively
correlated with increase in sattva, and decrease in stress, depression,
rajas and tamas, from pretest to posttest, and from posttest to follow-up.
The
main hypotheses of this experiment were that chanting the Hare Krsna maha
mantra will increase the effects of sattva, and will decrease stress,
depression, and the effects of rajas and tamas. Secondary
hypotheses were that gender and age will not effect the variables mentioned
above. More specifically, it was hypothesized that, when controlled for the
effects of gender and age, chanting hare krsna will continue to effect
the five dependent variables in the manner described above. A related research
hypothesis was that the frequency with which one chants the Hare Krsna maha
mantra will be positively associated with the strength of the effects on the
dependent variables.
Group
status was the primary independent variable being tested. Other independent
variables for which effects were statistically calculated were gender,
compliance with required chanting frequency, and age. Gender and age were
control variables. Dependent variables for this experiment included stress,
depression, sattva, rajas, and tamas.
Chanting
was a categorical variable, with three categories corresponding to the three
groups- maha mantra, alternate mantra, and control. These groups will be
described in greater detail in the procedures section. Gender was a categorical
variable, and chanting frequency and age were treated as interval variables,
with age measured as an integer and chanting frequency calculated as a fraction
of complete compliance with the instructions of the experiment. Measurement of
age and chanting frequency are described in more detail in the measurement
section. Instruments for measurement of the dependent variables are mentioned
later in this introduction, and described more extensively in the measurement
section of the chapter.
There
are three gunas described in the Vedas- sattva, rajas, and
tamas. Descriptions of each guna were
compiled from the Bhagavad-gita As It Is, Chapters 14, 17 and 18 (Prabhupada,
1972). Sattva guna is characterized by qualities such as cleanliness,
truthfulness, gravity, dutifulness, detachment, discipline, mental equilibrium,
respect for superiors, sharp intelligence, sense control, and staunch
determination. Attributes of rajas guna include intense activity, desire
for sense gratification, little interest in spiritual elevation, dissatisfaction
with one’s position, envy of others, and a materialistic mentality. Qualities
associated with tamas guna include mental imbalance, anger, arrogance,
depression, laziness, procrastination, and a feeling of helplessness. The
gunas were measured with the Vedic Personality Inventory (Wolf, 1998) (see
Table 1).
Personal
stress was measured with the Index of Clinical Stress (ICS) (Abell, 1991) (see
Table 1), which assesses a person’s subjective response to external demands,
as opposed to measuring one’s life events as an indicator of stress level.
Depression was measured with the Generalized Contentment Scale (GCS) (Hudson,
& Proctor, 1977) (see Table 1). This is a short-form measure of nonpsychotic
depression specifically developed for repeated administrations (Hudson, &
Proctor, 1977). Table 1 shows the dependent variables and the instruments
proposed to measure them.
Table 1
Dependent Variables and Their Measuring Instruments
Dependent Variable
Instrument
Sattva
Vedic Personality Inventory
Rajas
Vedic Personality Inventory
Tamas
Vedic Personality Inventory
Personal Stress
Index of Clinical Stress
Depression
Generalized Contentment Scale
In
summary it was predicted that the chanting treatment group, compared with the
alternate mantra and control groups, will increase in sattva, and will
decrease in stress, depression, rajas and tamas. Chanting
frequency was expected to be positively associated with the strength of effects
of chanting Hare Krsna. To specify, increased chanting of the hare krsna maha
mantra was predicted to increase sattva, and to decrease stress,
depression, rajas and tamas. Further, as support for the main
hypotheses, it was hypothesized that gender and age will have no effect on the
dependent variables.
Design
This
study incorporated an experimental design, utilizing an experimental, control
and placebo group. Each group was pre- and post-tested with a package of
surveys, including a scale on depression, personal stress, sattva,
rajas, and tamas. Random assignment was used to place subjects in
groups. There was also follow-up testing for all groups.
Two
research assistants were employed for this study. Before contacting any
subjects, each research assistant was trained in the basic principles and
techniques of scientific experimentation, and specifically in the methods of
this study. Specifically, the research assistants received instruction in
confidentiality, preserving the integrity of data, and maintaining neutrality,
as well as in the paperwork and timeframe for this study, and the method for
teaching subjects how to chant.
For
pretest, posttest and follow-up there was a 5-day window for all members of all
groups to be tested. Ideally, to minimize history effects, all subjects should
have been tested at precisely the same time. This was not practically
achievable, however, and therefore a 5-day period of testing was used. This will
be further explained in the procedures section of this chapter. In addition to
the 5-day window for evaluation of all participants, each participant had an
individual 3-day window for completion of each testing phase. For instance, if a
subject was pretested on November 17, then posttest should occur, based on a
scheduled 28-day intervention, on December 15. Due to scheduling difficulties,
it was not always possible to schedule posttest after exactly 28 days.
Therefore, the participant was allowed to be tested within the 3-day window
December 14-16. A similar 3-day window was used for follow-up scheduling. Though
this windowing method increased the chance that history will adversely affect
internal validity, it was used to minimize diffusion effects. That is, it was
desirable for study participants to not know the identity of other participants,
for this might lead to discussion of intervention effects during the study.
Therefore, the design did not arrange for all subjects to congregate together
during any phase of the study.
After
pretesting, members of the control group were informed that they were the
control group, and they were requested to return in 28 days for the posttest. A
member of the research team phoned control group members one or two days prior
to posttest, to remind them of the appointment. Members of the treatment group
were taught the intervention after pretest, and members of the placebo group
were taught the placebo intervention. These two groups were instructed how to
chant, and directed to do so for the next 28 days. Within three days after
pretest for a member of a chanting group, a member of the research team phoned
the chanter to check if there were any difficulties with following the
procedures.Seven or eight days after a subject initially learned the chanting,
the research team member met in-person with the subject to ensure that he/she
was properly following the procedures. After this meeting, a member of the
research team phoned the chanters approximately every seven days till the end of
the chanting phase of the study, to ensure that procedures were properly
followed. For these mid-intervention meetings and phone calls there was also a
5-day window, as well as a 3-day window for checking each individual
participant, as described above with regards to posttest and follow-up. Each
member of the study was phoned one or two days before follow-up, to remind them
of the appointment.
Intervention
The
subject was given a string of 109 beads (japa beads), with one bead
markedly larger than the others. The research team member had his/her own set of
japa beads, and demonstrated the chanting method for the subject. With the
thumb and middle finger of the right hand, the chanter holds the bead on either
side of the large bead. Then the subject chanted the mantra. After the
mantra is completed, the subject moves one bead through the fingers so that s/he
is holding the second bead from the large bead. Again the chanter should chant
the maha mantra. In this way, the participant should chant one mantra for
each bead, until 108 mantras have been chanted. This constitutes one “round”
of japa meditation. Japa can be performed in any circumstance. For
instance, one may be sitting or walking. The essential factor is that one is
fully attentive to the chanting. The treatment group was taught to chant the
maha mantra- hare krishna hare krishna krishna krishna hare hare/ hare rama
hare rama rama rama hare hare, and the alternate mantra group was taught to
chant sarva dasa sarva dasa dasa dasa sarva sarva/ sarva jana sarva jana jana
jana sarva sarva. The sarva dasa pseudo-mantra is a meaningless
combination of Sanskrit syllables made up by the author. The alternate mantra
was composed of the same syllabic pattern as the maha mantra. This
controlled for the effects of syllabic pattern, and helped to isolate the
effects of the mantras themselves.
Members
of both groups were instructed that this was an experiment on the efficacy of
chanting the particular combination of words that they learned to chant. After
the first day, subjects chanted on their own, though the research team member
informed subjects that they could contact him/her if there were questions or
concerns. Also, as described above, a research team member checked on the
chanters approximately every seven days for the course of the intervention
period. After 28 days of chanting, all three groups were posttested, and 28 days
after posttest, all three groups were follow-up tested. Researchers kept a log
of client contacts.
Internal Validity of Design
Internal
validity of a design refers to the design’s ability to allow inference about
causality regarding the relationship between measured dependent and independent
variables. Threats to internal validity of a design include history, maturation,
testing, instrumentation, statistical regression, selection, mortality,
diffusion, compensatory effects, resentful demoralization of subjects, and
various interactive threats, such as the interaction between selection and
maturation, selection and history, and selection and instrumentation (Cook and
Campbell, 1979).
Random
assignment, as used in the experiment described in this dissertation, reduces
threats to internal validity. Of course, randomization is probabilistic, and
does not ensure equivalence of groups.
The
threat of diffusion was not eliminated with random assignment, and could be a
factor in this study. However, each subject was dealt with privately, not in a
group, except for a few instances when roommates or friends knew of each
other’s involvement in the experiment. Still, most of the dozens of
participants resided in the same mid-sized town, and therefore it is possible
that subjects discovered the identity of each other, and they may have discussed
their experiences of the intervention. Thus diffusion of treatments may be an
issue. Research team members cautioned participants about discussing the study
with other subjects, in case the identity of other subjects become known to
them. Another threat to internal validity that may be active in this study is
resentful demoralization of subjects placed in a group perceived to be less
desirable. For instance, members of the control group may have become
discouraged because they didn’t participate in the intervention. Posttest
scores may reflect this discouragement, and this compromises the internal
validity of the design. Further, while random assignment may effectively deal
with the selection-mortality threat, mortality in this study did differ by
group. The placebo mantra group experienced greater attrition than the maha
mantra group, and this may be considered a threat to validity. However, such
a mortality effect can also be viewed as a predicted consequence of the group
design, especially when randomization is present (Cook & Campbell, 1979).
Cook
and Campbell (1979) differentiate between internal validity and construct
validity, with threats to construct validity conceived as effects that can lead
to invalid conclusions about the construct labels that should be attached to
manipulations and measures. Construct validity threats include the effects of
the experimenter as an independent variable, and the Hawthorne effect. Random
assignment with a placebo intervention group should control for a Hawthorne
effect, though differential effects of contact between research team members and
the members of various groups may be a threat to internal validity. To reduce
this risk, each team member was instructed in the importance of uniformity when
they present study instructions and techniques to participants in the study.
Sampling
Participants
in the study were self-selected, and then screened by the researcher. The
researcher placed advertisements in the college and city newspapers of a major
state university in a mid-sized town in the Southeastern United States,
requesting volunteers for this experiment. The researcher chose the first 93
acceptably screened respondents to the newspaper advertisements. Respondents
were screened by phone interview by the researcher for traits such as
debilitating mental disorders, availability to participate in the entire
duration of the study, willingness to devote approximately half an hour a day
for involvement in the study for four weeks, and prior experience with chanting,
biofeedback, meditation, and other yoga techniques. Information on prior
experience of subjects with yoga, biofeedback, meditation, and chanting was used
to create two blocks, “past experience” and “no past experience”, for
the random assignment to groups. With this method, the factor of prior
experience was controlled. At the pretest meeting each subject signed a consent
form for participating in research.
On
the morning following the final day of newspaper advertisements, names of
participants were assigned numbers, and participants were assigned into groups
with a random number table. Specifically, the participant whose number corresponded with
the first number chosen from the random number table was placed in the maha
mantra group, the participant whose number corresponded with the second number
chosen was placed in the control group, the participant whose number
corresponded with the third number chosen was placed in the placebo group, the
participant whose name corresponded with the fourth number chosen was placed in
the maha mantra group, and so on. As described above, two blocks of
participants, those having experience with mantras, meditation, yoga, or
biofeedback, and those without experience, were randomly assigned to groups.
All
participants were at least 18 years of age and signed a form of voluntary
consent before participating. Thirty-three subjects per group began the study.
To facilitate recruitment a $40 financial incentive was offered to participants
who completed the study. Though such monetary encouragement may have affected
the type and motivation of subjects, the strategy of random assignment was used
to equally distribute such effects in all three groups.
External
Validity of Design
According
to Vedic theory, the effects of chanting hare krsna on the chosen
dependent variables should apply to all people. Since the sampling method of
this experiment was not random, results of the study cannot be inferred to apply
to any population or sampling frame. Thus, the lack of random sampling greatly
diminishes the generalizability, or external validity, of this study.
Bracht
and Glass (1968) distinguish between population and ecological validity, both of
which are forms of external validity. Ecological validity refers to the
conditions, including settings and experimenters, under which the same results
of an experiment can be expected, and population validity refers to the persons
to whom the results can be applied. Ecological validity assumes that the
experimental effect is independent of the environment. With the sampling
strategy of this experiment, it is not possible to assume that the effects will
apply in different environments. For example, this study was conducted in a
Southeastern university town, and therefore the results may not apply to a
setting such as a Northeastern industrial city, or a rural area of a third world
country.
For
population validity, the experimenter must make two inferential leaps- from the
sample to the experimentally accessible population, and from the accessible
population to the target population. Since the sample was self-selected, rather
than randomly selected, even the leap to the population of the county where the
study occurred is not possible.
Other
factors affecting external validity in this study include novelty effects,
experimenter effects, and pretest sensitization. Novelty effects refer to the
effects caused by the newness of an intervention, which will fade after some
time. These effects reduce the generalizability of the findings, as do effects
caused by interaction between the experimenter and subjects. Pretest
sensitization refers to the possibility that response to the treatment may be
affected by the pretest, and therefore results are not generalizable to persons
who did not take the pretest.
Though
the external validity of this design is not strong, the procedures of the
experiment are specified, and replication of the study is possible. With
replication in various settings and with diverse populations, generalizability
of the results of this study can be confirmed or disconfirmed.
Measurement
The
instruments that were used in this study were the Vedic Personality Inventory (VPI)
(Wolf, 1998), the Generalized Contentment Scale (GCS) (Hudson, & Proctor,
1977), and the Index of Clinical Stress (ICS) (Abell, 1991). A description of
the VPI was given in the measurement of spirituality section of the previous
chapter, and a summary of that description is provided below. All of the
instruments used in this study are found in Appendix D.
The
Vedic Personality Inventory
The
VPI is an instrument that measures the three gunas- sattva,
rajas, and tamas. The instrument contains 56 items, each of them with
seven Likert-type response choices. Internal consistency alpha for the three
subscales ranges from .93 to .94. Evidence for construct validity for the
instrument has been obtained with relation to measures on verbal aggressiveness,
hours of sleep per day, and life satisfaction. Regarding factor analysis, all
items correlate positively and significantly with their intended subscale.
The
Generalized Contentment Scale (GCS)
The
GCS (Hudson & Proctor, 1977) measures the magnitude of nonpsychotic
depression, and consists of 25 items. It is a summated category partition scale
that is scored on a range from 0 to 100. Positively and negatively worded items
are used to counter response bias. Psychometric testing of the GCS was done on a
sample of 124 persons of diverse occupations, and income and education levels.
Test-retest and split-half reliability scores for the GCS ranged from .887 to
.963, with a mean of .930. Also, the instrument showed good ability to
differentiate between groups who described themselves as depressed and not
depressed. Construct validity for the GCS has also been established.
According
to Hudson (1982), the GCS not only possesses strong reliability and validity, it
is also suited for repeated measures with the same client. Specifically, the GCS
is short, easy to complete and score, and does not suffer from response decay
when used repeatedly over time. For these reasons, the GCS has been chosen as a
measure for this pretest, posttest experimental design. Hudson (1982) provides
scoring procedures for the instrument, including procedures to use when the
subject does not complete all items. These procedures were used for scoring the
GCS in this pilot study.
The
Index of Clinical Stress (ICS)
This
measure was designed to assess the subjective aspect of stress in a generalized,
unidimensional form (Abell, 1991). The 25 items of the instrument were designed
to reflect the range of perceptions involved with subjective stress. This
approach to measuring stress is differentiated from stress measurement as a
result of external life situation and life events. An internal approach to
assessment of stress level is appropriate for japa intervention, since chanting
is hypothesized to positively alter one’s consciousness, or internal state.
Each
ICS item has a five-point response range, with some items negatively worded to
avoid response bias. Abell (1991) provides scoring procedures for the scale,
including procedures to use when not all items are completed.
Psychometrics
of the instrument were assessed using a sample of 265 persons, whose mean age
was 33 years. Cronbach’s alpha for the sample was .963. The ICS has also shown
strong factorial validity, convergent construct validity, and discriminant
construct validity. Based on these results, Abell (1991) suggests that the ICS
can be used with confidence by social work practitioners and researchers when
single or repeated measures of subjective stress are required.
Compliance
Log
Subjects
in the chanting groups were required to chant three rounds of japa per
day. At pretest they were given a compliance log, in which they recorded the
number of rounds they chanted each day for the duration of the intervention. A
copy of this Compliance Log is in Appendix D. After the followup period a
compliance percentage was calculated. The compliance percentage was calculated
as:
number
of rounds chanted during intervention x 100
number
of days in intervention period x 3
For
example, if the subject chanted for 28 days, then 100% compliance would be 28 x
3 rounds = 84 rounds: (84 x 100)/(28 x 3)
= 8400/84
= 100%
Other
Data
During
the initial phone conversation with each subject, information was gathered
regarding prior experience with chanting, biofeedback, meditation, and other
yoga practices. Also, each member of the research team noted comments after each
contact with a subject, paying special attention to factors such as anxiety
about receiving payment for study, enthusiasm for the chanting, and
discouragement about participation in the study. Experiences with the subjects
of the pilot study alerted the researcher to the possible importance of noting
reactions of these types.
Procedures
From
November 5-14, 1998 (see Table 2), the researcher placed newspaper
advertisements in three community and student newspapers in a university town in
North Florida. The advertisements read as follows:
Wanted:
Participants in a study examining the effects of an Eastern-style intervention
on stress and depression. $40 reimbursement. If interested, contact David Wolf
at [phone number].
When people
responded by telephoning the researcher, he explained to them that this is an
experiment testing the efficacy of an Eastern-style intervention on factors such
as stress, depression, and spirituality. One-hundred-eight calls in response to
the ads were accepted. Of the 108 callers, 93 made it to the stage of being
randomly assigned to an experimental group, and the others either didn’t
respond to return calls, or lost interest after hearing about the study. For the
93 participants, pretest meetings were scheduled from November 16-20, 1998 (see
Table 2), at a place where the respondent felt comfortable, such as their home
or a public place on campus. Also, during the phone call the researcher asked
each subject “Do you have prior experience with chanting mantras, meditation,
yoga, or biofeedback?” and noted the answer. Results from this question were
used to create two blocks (“past experience” and “no past experience”)
for the random assignment. On November 15, 1998, the researcher, by use of a
random number table, randomly assigned respondents from each block to the three
groups- maha mantra, alternate mantra, and control.
From
November 16-20 the researcher, or one of two research assistants, met with the
study participants for a pretest meeting. At the beginning of each meeting the
research team member explained the nature and purpose of the research, and the
subject read and signed the Consent Form for Participating in Research. After
signing this form each participant completed the package of surveys, including
the VPI, ICS, and GCS. At the bottom of the VPI the participants filled in
demographic items for gender and age. For control group members the research
team member scheduled a posttest meeting between December 14-18, 1998 (see Table
2), making all efforts to schedule the posttest exactly 28 days after the
pretest. For members of the two chanting groups, the research team member
instructed the subjects to chant on japa beads, as described in the Design
section of this Methodology chapter. At the end of the pretest session the
research team member paid each subject $20. Upon receipt of this payment each
subject completed a payment receipt (Appendix D). Also, at the end of the
pretest session subjects in either of the chanting groups received a compliance
log (Appendix D), and the research team member instructed them to record in this
log the number of japa rounds they chanted each day. Control group members were
informed at pretest that at the end of the study they would be introduced to the
mantra.
Of
the 93 subjects who were randomly assigned to the three groups, 81 completed the
pre-test session. Regarding the other 12, they either didn't show up for the
pre-test meeting and didn't respond to attempts to schedule another meeting
within the 5-day window, or during the pre-test meeting they decided they didn't
want to participate.
Each
subject was phoned by a research team member one or two days after pretest, to
ensure that they understood and were following the procedures of the experiment.
A researcher visited each member of the two chanting groups between November
23-27. This visit occurred 6 to 8 days after pretest. The purpose of this visit
was to ensure that the participants were chanting properly. During the visit the
research team member chanted at least one round with the subject. Also,
compliance logs were collected, and new compliance log sheets were distributed
to subjects. From November 30-December 4, a research team member phoned each
subject to check if there were any difficulties in participating in the study,
to check if procedures were being properly followed, and to obtain any notable
responses to the study by participants. A similar phone contact was made between
December 7-11.
Between
the dates of December 14-18 (see Table 2) there was an in-person posttest
meeting. During this posttest meeting, subjects completed the packet of
measuring instruments after completing their chanting for that day. This
posttest day was the final day of chanting for the two chanting groups. Posttest
meetings were scheduled within 27-29 days of the pretest. Members of the control
group were phoned one or two days before posttest to remind them of the
schedule. Of the 81 participants who completed the pretest session, 64 remained
after the posttest session. Of these 64, 20 were in the control group, 20 were
in the alternate mantra group, and 24 were in the maha mantra group.
For
the 64 participants remaining after posttest, two had their 1-week meeting one
day after the closure of the 3-day window, and one had the 1-week meeting 2 days
after the closure of the 3-day window. For two subjects the posttest meeting was
held two days after the closure of the posttest window. After the followup
meetings, when data was analyzed, it was decided to retain the five subjects
described above because none of them had outlying scores on any of the dependent
variables.
All
participants were phoned on January 9 or January 10, 1999, to remind them of the
follow-up appointment. A research team member personally met with each subject
to administer the follow-up surveys and pay the remaining $20. In some cases,
where a personal meeting was not possible, participants mailed the surveys to
the researcher in a stamped envelope provided by the researcher at posttest, and
the researcher mailed $20 to the participant upon receipt of the follow-up
package. Follow-up measures were completed within 27-29 days of posttest.
Participants signed a receipt for the $20 follow-up payment. Sixty-two subjects
completed the entire study. After all subjects completed the follow-up session,
surveys were scored and data was entered and analyzed using SPSS.
Table 2
Schedule for Hare Krsna Maha Mantra Japa Research Project
Newspaper Ads for Participants
November 5-14, 1998
5-day pretest window
November
16-20
5-day 1st week chanting check window
November 23-27
5-day 2nd week phone call chanting check window
Nov. 30- Dec. 4
5-day 3rd week phone call chanting check window
December 7-11
5-day posttest window
December 14-18
5-day follow-up window
January 11-15
Analysis
Data
for all instruments were treated as interval. For each instrument, a mean for
each group in the study was calculated at pretest, posttest, and follow-up. Mean
compliance log percentages were calculated for the two chanting groups, and mean
ages were calculated for all three groups. Thus, there were four independent
variables. Two of the independent variables, gender and group status (maha
mantra, alternate mantra, or control) were categorical, and two were interval
(age and compliance log frequency). There were five dependent variables, which
were measured at the interval level by the instruments assessing stress (ICS),
depression (GCS), sattva (VPI), rajas (VPI), and tamas (VPI).
For
each of the standardized instruments there were three means (pretest, posttest
and follow-up) for each of the three groups. Therefore, there were 45 means for
the standardized instruments. Also, each of the two chanting groups had a mean
compliance log frequency, and all three groups had an average age.
The
main effects examined in this study were the effects of the group (maha
mantra, control, and alternate) independent variable on the five dependent
variables. Analysis of data in this study utilized the hierarchical analysis
approach described by Cohen and Cohen (1983). With this approach to data
analysis, independent variables are entered cumulatively in a prespecified
sequence, in accord with the purpose and logic of the research. This procedure
permits extrication of as much causal inference as the data allow. This is
accomplished by protecting the main effects from potentially weakening
statistical influences produced by secondary variables (Cohen & Cohen,
1983). Applied specifically to this study, gender and age were analyzed at the
first level, since they are not caused by the group independent variable or
chanting frequency. Since gender and age could be spurious variables in the
analysis of the chanting variables, it was best to remove their effects before
analyzing the variables related to chanting. Chanting frequency was analyzed at
the next level of the hierarchy, and then the group status variable was
analyzed. ANCOVA, with gender, age, and chanting frequency as covariates, was
used to determine similarity of group means. When ANCOVA resulted in a
significant F test, t tests were performed to determine which groups
significantly differed. Analytical procedures will be described in more detail
in the following chapter.
Power
Analysis
Power
analysis for this 3-group design is explained in Cohen and Cohen (1983). Alpha
level for each F test was .05. Orme and Combs-Orme (1986) offer three general
strategies for determining effect sizes. These are estimations from previous
research, effect sizes selected to indicate a population effect that would have
either practical or theoretical significance, and effect sizes chosen from
suggested conventional definitions of small, medium and large effect sizes.
Based on the study of Janowiak and Hackman (1994), the maha mantra
single system design pilot study, and Vedic theory, an effect size of at least r2
= .3 was expected. To be reasonably sure that effects would be detected, an r2
of .25 was used for calculation of power and sample size. According to Rosenthal
(1997), an r2 of .25 indicates a strong association and large effect size.
Multiple R2, the effect size of all
the independent variables combined, was estimated to be .50, with chanting
frequency expected to account for most of the explained model variance not
explained by group status. From formula 4.5.1 on page 155 of Cohen and Cohen:
f2
= (R2Y.AB - R2Y.A)/(1-R2Y.AB)
= (.50
- .25) / (1 - .50) =
.25/.50 =
.50
Table
E.2 on page 527 (Cohen & Cohen, 1983) provides power values for alpha = .05.
Referring to that table, kB, which is the number of maain effect independent variables,
equals one. Therefore, from table E.2, to achieve power of .8, as recommended by
Cohen and Cohen, the L value for kB =
1 is 7.85.
From
page 155, formula 4.5.2, sample size (n) =
L/f2 + kA
+ kB + 1. kB
= 1, as described above, and kA
= 3. kA represents the number of
covariates.
Therefore, n = (7.85/.50) +
3 + 1 + 1 = 20.70
Thus,
to achieve a power of .8 (Type II error rate of .2), with alpha (Type I error
rate) = .05, a sample of at least 21 persons per group is necessary to detect an
effect size for group status of r2 =
.25.
Statistical
Tests
An
ANCOVA was performed for each of the five dependent variables, and the t-tests
within the ANOVAs were protected t tests (Cohen & Cohen, 1983). Using the
protected t-test combines strong power attributes of the individual t test with
the protection against large experiment-wise Type I error, due to the condition
that the F test must meet the alpha significance level in order for subsequent t
tests to be performed. Thus, the t tests are protected from the accumulation of
small alpha levels to large expermimentwise error rates. Cohen and Cohen
conclude that the protected t test is effective in keeping Type I errors low
while maintaining good power levels. The independent variables of gender and age
underwent the same procedures for analysis as the chanting-related variables,
although these two other independent variables were analyzed at the first level
in the hierarchical analysis in order to remove their effects from the
chanting-related variables.