You are on page 1of 28

Qaapteff 3

METHODOLOGY

rhe population - The sample — l l i c


variables - The tests — Procedure used for
data collection — ITie statistical tcchniqucs
used for analysis o f data.
The methodolog}' or design o f the present investigation can be
summarised into six heads; (1) the population, (2) the sample, (3) the
variables, (4) the tests, (5) procedure used for data collection, and (6) the
statistical techniques.

1. THE POPULATION

The population under study consisted o f adolecent students, both


boys and girls in the age group o f 13 - 21 years. There are many different
contemporary definitions o f adolescence, each o f which reflects a
somewhat different fecet o f the interaction between the society and the
adolescent. Medically, adolescence begins with the growth and hormonal
changes associated with sexual maturity and ends when there is no further
growth. Legally, adolescence ends with the asumption of adult
responsibility for voting, for the draft, and for debts and contracts,
including marriage. Age for the assumption o f these duties range firom 14 to
21 years in various countries. There are also variations in the age at which a
child bccomes responsible for criminal acts : these range from 14 to 18 in
most countries (I laviland & Scarborough, 1981).

Psychologically, adoloscence is a period o f transition during which


physical, cognitive, affective, and behavioural changes occur, culminating in
the practce o f adult roles and the formation o f personal identity (Santrock,
1996).

In every case, we can see that the definitions depend on various


contemporary perspectives on the physical and social criteria for adulthood.
I'here arc no absolute age limits or concrete boundaries, exccpt for those
imposed by the law.

Western researchers have included ciiildren from ten and a half


onwards as adolescents. Hurlock (1967) dates the beginning o f adolescence
from ‘sexual maturing’ which she obseves on the average to the thirteenth
year o f life for the girls and the fourteenth year for boys.

It must be recognized that any attempt to affix specific ages to


these physical developments is at best inexact. Chronological age is a poor
indicator o f biological age and especially so at adolescence because o f the
great individual differences. Though for girls menarche is generally
considered as the identifying event o f puberty, for boys any such event is
altogether absent. The appearance o f secondary sexual characteristic as
th a t

visible marks o f puberty is complicated by the fact^there are considerable


variation in the intervening periods between the appearance o f spermatozoa
and the external manifestations o f visible bodily changes. For the present
investigation thirteen years is taken as the beginning o f adolescence both
for boys and girls.

For the end, as for the beginning o f adolescence, both the choice o f
cdteria and the tremendous individual variations in meeting them
complicate efforts at age definition. Different researchers have fixed
different ages ranging from twenty to twenty five years as marking the end
o f adolescence. Hurlock (1967) prefers to regard twenty one as the end o f
adolescence, i^or die present investigation also twenty one years is taken as
the end o f adolescence.

Developmentalists generally describe adolescence in terms o f early


and late periods, f^^arly adolescence roughly covers pubertal change,
whereas, late adolescence covers approaximately the latter half o f the
second decade o f life (Santrock, 1996). Researchers who study adolescencie
increasingly specify whether their results are specific to early or late
adolescence.

Konopka (1973) distinguishes early adolescence (the years from


twelve to fifteen), middle adolescence (fifteen to eighteen), and late
adolescence (nineteen to twenty two). For the present investigation a
classification similar to the above one is adopted, dividing the adolescence
into early, middle, and late periods,except that the onset is taken as thirteen
years and the completion as twenty one years.
In the present study, the nine year period covered by adolescence is
divcded into three stages o f development, viz., early adolcsccnce, mid­
adolescence, and late adolescence so that we can have a closer look at the
rapid changes taking place in this period. Since there will be considerable
overlap in the mid-adolescence, separating that stage will provide an
opportunity to clearly bring out the conspicuous differences between the
onset and culmination o f adolescence.

Since family environment is embedded in a larger socio-cultural


mileu, urban-rural difference is expected to play an important role in
determining family climate.

Only student population was included in the study based on two


considerations:

(1) Education is universal in Kerala and those outside it is only


marginal; and
I
•V

(2) instruments are designed to test the literate subjects.


i/

But, students b e lo n ^ g to all kinds o f institutions like schools,


colleges, iT ls, polytechnics, professional colleges, tutorials, and parrallcl
colleges were covered.

2. THE SAMPLE

'I’he sample was drawn from the student population o f Kerala, from
different kinds o f institutions including schools, colleges, techmcal
institutions, professional colleges, tutorials, and parallel colleges, located at
both urban and rural areas. Adolescents o f both genders were included in
the sample. A stratified random sampling procedure was employed to select
the sample.

Out o f an initial sample o f 629 subjects selected foe the study, 592
subjects were taken for the final analysis. The data sheets from the
remaining 37 subjects were found incomplete and were rejected.

Out o f the 592 subjects, 282 were boys and the remaining 310 were
girls. These included 188 early, 204 middle, and 200 late adolescents. O f
these 354 were rural and 238 were urban subjects.

I'hese includc 164 subjects studying in nincth and tenth classes in


seven schools, o f which three were situated in the urban area and four were
situated in rural areas, A total o f 148 subjects studying for Prcdegree, and
134 subjects studying for Degree courses were selected from six colleges,
three from the city and three from the suburbs. A group o f 76 subjects
were selected from professional institutions including Engineering College,
Medical College, Polytechnic, and Ayurveda College. The remaining 70
subjects were drawn from parallel institutions like tutorials, parallel colleges,
vocational training centres etc. "Hiey also belonged to different socio­
economic and religious segments o f the society ( see Appendix A).

3. THE VAIUABLES

The variables selected for the present study can be grouped under
two broad heads; family climate and mental health.
FAMILY CLIMATE VARIABLES:

Family climate is treated as a global construct that describes the


psychological-'environment prevailing in the fan>ily, comprising o f six
variables. A review o f literature shows that though different researchers
have studied different aspects o f the family environment, there are few
studies that include all the aspects o f the environment within a
comprehensive framework. The long list o f variables can be conceptually
incorporated into six variables representing the family environment —
paternal control, maternal control, paternal affect, maternal affect, parental
harmony, and sibling relationships.

1. Paternal Control:

Different researchers used different labels to conccptualize and


measure parental behaviour. These included discipUnary styles (Hurlock,
1978); restrictivencss (Baldwin, 1955; Baumrind & Black, 1967; Becker,
1964); firm control and psychological control (Berger & Armentrout, 1971
Schacfer, 1965); demandingness (Roe & Siegelman, 1963; Siegelman, 1965)
authoritative, authoritarian, and permissive (Baumrind, 1967, 1971)
directive/conventional, assertive, and supportive control (Baumrind, 1991a,
1991b); overprotection (l^arker, et al., 1970); coercive, inductive, and
unspecified control (Rollin & I'homas, 1979); power-asscrtivencss (I’arke &
Slaby, 1983); etc.

I ’he present study instead o f treating parental control as a


single variable, paternal and maternal controls were treated as separate
variables. Paternal control is defined as the the way in which the father
exercises authority over his children. Operationally, control is defined as
power-oriented, highly demanding, placing high value on
conformity/obedience, resorting to force and punishment to deal with
disobedience, and biased towards parental needs.

2. Maternal Control

Maternal control is defined as the way in which the mother


exercises authority over her cliildren. Just like paternal control, maternal
control is operationalized as power-oriented, highly demanding, placing
high value on conformity/obedience, resorting to punishment to deal
with disobedience and biased towards parental needs.

3. Paternal Affect:

As in the case o f control, parental affect has been the subject o f


several earlier studies, using different labels such as attachment
(Ainsworth, 1979; Bowlby, 1969, 1989; Santrock, 1995); connectedness
(Allen & Kupperminc, 1995; Black & McCartney, 1995); warmth and
nurturance (Metherington & Parke, 1993); emotional expressiveness
(Boyum & Parke, 1995; Cassidy^et al., 1992; Denham, 1993); etc.

I'or the present investigation, parental affect is differentiated


into paternal affect and maternal affect. Paternal affect is measured in
terms o f the extent o f the father’s attachment, expression o f affection,
and sensitivity to cliildren’s needs.
4. Maternal Affect:

Just like paternal affect maternal affect is also measured in terms


o f mother’s attachment, expression o f affection, and sensitivity to her
children’s needs.

5. Parental Harmony:

Previous researchers have conceptualized marital relationship


focusing on the maladjustment dimension leading to children’s
behavioural problems. Marital conflict ((Carlson, 1984; Cummings &
Davies, 1994; Kreig, 1995), intraparental aggression (Gordis^et al.,1977)
etc., arc examples. I'or the present investigation the positive dimension
o f marital relationship is taken so that it can be directly correlated with
children’s mental health. Parental harmony is defined and
operationalized as the extent o f mutuality througli integration o f needs
and interests, trust and support, and sharing o f responsibilities o f both
parents.

6. Sibling Relationship:

'I'here have been extensive research on the impact o f sibling


relationship on the behaviour o f children. An individual’s social skill
development is influenced by the early interaction patterns with the
sibling. So any study on family climate has to include sibling
relationship as an antecedent variable influencing behavioural outcomes
o f children. In the present study, the sibling relationship is defined as
the extent o f interaction, intimacy and influence among siblings.
MENTAL HEALTH VARIABLES:

Mental health is the balanced development o f the individiial’s


pctsonality and emotional attitudes which enables him to live
harmoniously with himself and his fellow men. Though there are a long
list o f different attributes identified by different researchers, there seems
to be a certain degree o f agreement among psychologists on the
following four variables as most essential to mental health.
I
1. Self-Esteem:

Researchers with a social psychological orientation have pointed


out that (iur self-conecpdons emerge from the interactions with
othersand that our self-conceptions influence and guide our behaviour
(Cooley, 1902,1909; Mead, 1934; Sullivan, 1947, 1953). There are
several studies investigating self-esteem both as an outcome o f family
interactions as well as an antecedent to many behavioural outcomes. In
the present investigation, self-esteem is defined as the sense o f self-
worth or self-image referring to a person’s positivfc and negative self-
evaluations.
<r‘

2. Autonomy: ^

Steinberg and Silverberg (1986) have conceptualized autonomy as a


multi-dimcnsional construct, the three dimensions o f which are emotional
autonomy, resistance to peer pressure, and subjective sense o f self-reliance,
llie development o f autonomy has impotrant implications during
adolescence because it causes radical shifts in the relationship patterns. It
has been a an area o f a great deal o f research among developmentalists. For
the present investigation, autonomy is defined as the sense o f self-direction,
self-initiation, increased ability to make independent decisions, and
emancipation from undue social influence.

3. Emotional stability:

Adolescence is considered to be a period o f heightened


emotionality. One o f the basic developmental dimensions is emotional
maturity. So for the present invcsigation emotional stability is taken as a
mental health variable. Emotional stability is defined as the capacity for
positive and spontaneous emotional experience, steady and balanced
emotional expressions, and control over unrealistic and disruptive
emotions.

4. Social competence:

There is a ^ w i d e variety o f definitions stressing on diverse


outcomes and skills. According to Durkin (1995) socially competent
individuals are expected to be able to communicate effectively, to
understand the self and others, to acquirc a gender role, to obsefve the
moral order o f their community, to regulate their emotions, to adapt their
behaviour to age related norms, and so on.

For the present investigation, social competence is operationalized


as the capacity to develop and sustain intimate, harmonious, and
meaningful relationships with mutual acceptance and trust, recognizing
each others individuality and purpose.
4. THE TESTS

I ’wo instruments were constructed for the present study — Tamily


Climate Scale’ to measure the six family climate variables and ‘Mental
Health Profile’ to measure the four mental health variables.

1. Prepafation of the Draft Scales

a. Family Climate Scale:

A detailed review o f previous studies show that there has been very
few comprehensive family climate instrument measuring various
dimensions o f family environment, "nie I'amily Environment Scale (PF.S)
developed by Moos, et al. (1974) was employed to assess dimensions o f
overall family climate. ITiis consists o f 90 true-false items and yielded 10
subscale scores: Cohesion, Expressiveness, Conflict, Independence,
Achievement Orientation, Active-Recreational Orientation, Intellectual-
Cultural Orientation, Moral-Religious Emphasis, Organization, and
Control. A number o f studies have independently measured different
family environment variables using vaious methods including interviews,
rating scales, check lists, self reports, observational categories, and so-on.

Simons^et al. (1994) used multiple methods to measure dimensions


o f parenting - aggregating parent’s self-reports, adolescent’s reports, and
obsever rating o f the videotaped family interaction tasks.

I ’here has been considerable controversy over whedier


environmental variables are to be measured as objective measures or as
perceptual measures. In a number o f areas o f applicational research, like
organizational behaviour, the present day investigators generally employ
perceptual measurement in climate research. According to Jones and James
(1979), such an approach presupposes a number o f basic assumptions:

that psychological climate: (a) refers to the individual’s


cognitively based description o f tlie situation; (b) involves a
psychological processing o f specific perceptions into more
abstract depletions o f the psychologically meaningful
influences in the situation; (c) tends to be most closely related
to situational characteristics that have relatively direct and
immediate ties to individual experience; and (d) is
multidimensional, with a central core o f dimensions that apply
across a variety o f situations (though additional, specific
dimensions might be needed to better describe particular
situations).

The present researcher also subscribes to this line o f thinking, and


family climate is operationalized as a perceptual variable.

There have been a variety o f instruments used to measure different


aspects o f parental behaviour, either as self-reports or as it is perceived by
the children. For example, Autonomy-Control Scale (de Man, 1987-88)
which consists o f 30 statements that concern child-rearing practices that
differ in amount o f control. Responses are indicated on five-point scales.

Parent Acceptance-Rcjection Questionnaire' (l^ARQ) developed by


Rohner (1989) consists o f 60 statements concerning parent behaviour
towards the child (e.g., My mother says many unkind words to me).
Subjects respond to each statement on a four-point scale. ITie measure
yields four subscalc scores; warmth, aggression, rejection, and neglect.
Family Adaptability and Cohesion Evaluation Scales II (FACES II:
Olson ct al., 1982) and Scale III (FACESIII: Olson, 1986) were designed to
assess family cohesion (i.e., the emotional bonding the family members
have towards each other), and adaptability (i.e., the ability o f the family to
change in response to situational and developemntal stress). O ’Leary-Porter
Scale (1980) measures marital and family fiinctioning. This scale was
originally designed to assess how overt marital hostility might be related to
child conduct problems, 'i’he scale includes 10 items (e.g.. How much do
you argue with your spouse in front o f your child?) which the parents rated
on a five-point Likert-typc scale.

Colorado Self-Report o f I'amily I'unctioning Inventory (Bloom,


1985) used a four-point Likert-type scale measuring family cohesion (5
items: e.g.. Family members really help and support one another), and
i
enmeshment (5 items: e.g., Family members feel guilty if they want to spend
some time alone).

Braikcr and Kcllcy (1979) developed a nine-item love scale -


Relationship Questionnaire - to evaluate marital relationship. Maxwell
(1985) also developed 14-item Close Relationships Questionnaire. '

Hazzard, et al.(1983) developed Parent Perception Inventory (PPI)


in which the children rated their parents on a five-point Likert scale on
dimensions o f warmth and criticism.

Bury (1989) developed the Parental Authority Questionnaire


(PAQ), to tap those dimensions of parental behaviour outlined by
Baumrind (1971). Conlflict I'actics Scalc (C l’S) was developed by Straus
(1979), to measure two factors o f family conflict, verbal and physical.

Bengston and Schrader’s (1982) Positive Affect Scale measured

parental understanding, trust, respect, fairness, and affection (e.g., How well
do you feel that your mother understands you?; How much do you trust
your father?). The items were rated on a three point scale.

Children’s Report o f Parental Behaviour Inventory (CRPBI) was


developed by Schaefer (1965) to assess: acceptance (8 items), rejection (10
items), inconsistent discipline (8 items), control (8 items), and hostile
control (8 items) on three-point response scalcs^

Self-Image Questionnaire for Young Adolescents (SIQYA)


developed by Peterson^ et al. (1984) consists o f a Family Relations Scale
(12 items). Items reflecting parental approval and warmth include: my
parents are ashamed o f me (reverse coded); I can count on my parents most
o f the time; I feel I have a part in making faihily decisions etc. Items
reflecting conflict and negative feelings about the home include: 1 try to stay
away from the home most o f the time; my parents are difficult to
understand; etc.

Sarason and Bell (1960) developed a Parents’ Attitude Scale (PAS)


based on an original scale developed by Schaefer and Bell (1958). Based on
these Malhotra (1990) developed a 14-item Parental Handling
Questionnaire (l^HQ) with two factors o f care and control. , •
Family Integration Inventory developed by Abraham and
Fernandez (1978) consists o f 60 true-false items measuring six family
variables: Home Environment, Home Relations, Accepatance-Rejection,
Dependence-Independence, Autocraq^-Democracy, and Mutual Trust and
Approval.

Sibling Relationship Inventory (SRI) was developed by Stoker and


McI lalc (1992) to evaluate sibling rclationsips.

Adapting items from the above scales and incorporating some


original items suitable for our social and cultural values and norms, items
for each o f the six family climate variables were framed. As in Family
Environment Scale (FES) by M oos,et al. (1974), and a number o f other
scales, ‘true-false’ response categories were given in order to avoid neutral
responses and to make the responses more quick and spontaneous. For
paternal and maternal control same items were used with separate
responses for father and mother. ITie same pattetn was used for paternal
and maternal affect as well.

An initial set o f 56 items were developed for the draft scale. Both
positive and negative statements were included. Exacriy equal number o f
positive and negative items could not be maintained, since some items
sounded natural and non-committal when phrased into positive statements
while some sounded natural in negative expression. The positive and
negative items were arranged randomly. Both English and Malayalam
versions o f die items were prepared.
b. Mental Health Profile:

As in the case o f family climate, no single and comprehensive


instrument is available to measure the four mental health variables chosen
for the study. Majority o f adolescent studies focussed on the development
o f behavioural problems rather than healthy or adjustive behavioural
patterns. But a number o f previous researchers have developed separate
scales to measure the four variables.

The Global Self-Worth Scale (Harter, 1985), Self-Esteem Scale


(Rosenberg, 1965), Self-Esteem Inventory (Battle, 1981), Self-Esteem
Inventory (I’homas & SanandaRaj, 1982), Emotional Autonomy Scale
(Steinberg & Silverberg, 1986), Harter’s Perceived Competence Scale for
Children (Harter, 1982, 1985), Rutter Child Scales (Rutter, et al, 1983)
Socio-Personal Adjustment Scale (Nair,1976) which is an adapted version
o f California Test o f Personality (CTP), are some o f the widely used tests in
this area. Items for the present instrument —Mental Health Profile —were
adapted from the above tests. A few items were originally framed by the
present researcher. An initial set o f 60 items were included in the draft
scale. Both English and Malayalam versions o f the questionnaires were
prepared.

2. Administration of the Draft Scale:


!

The tests were administered to aii initial sample o f 250 adolescents,


in the age group o f 13-21 years, both rural and urban, belonging to different
sociocultural groups, and studying in different types o f institutions.
including schools, colleges, technical institutions, tutorials, etc., using
stratified random sampling.

Subjects were personally contacted and instructions were read out


and clarifications were made. Out o f the 244 completed responses, 240
were selected.

3. Scoring and Item Analysis:

a. Family Climate Scale:

The scoring was done as follows: A score o f 1 was given to the true
and 0 for false in the case o f positive items. 'Fhe score was reversed in the
case o f negative items. A response sheet was not scored , if there were more
than one response category chosen for an item or if there were any omitted
item. While scoring, the total score obtained for all the items representing
each family climate variable was taken seperately. The total score for each
variable can range from 0 to 15.

Item analysis was carried out as per the Likert’s method (Edwards,
1969). I ’he t—values were estimated on the basis o f this analysis. The results
o f the item analysis show that out o f the 56 items used in the draft scale 53
have statistically significant t-values (at 0.01 level), indicating that all these
items have high discriminative power.

After arranging the significant items in the order o f their t-values,


ten items each for the six climate variables (for paternal and maternal
control ten common items; and for paternal and maternal affect ten
common items) were selected for the final test. Three critera were used in
selecting the items for the final test;

1. I'he item should have high t-values,

2. The items should represent a distinct aspect o f the variable, and

3. The item should be meaningful and applicable to jdl categories o f the


sample.

A total o f 40 items were thus selected for the final instrument. The
selected items were arranged in four different sub-scales, but the positive
and negative items were arranged randomly. The instructions were retained
commonly for all the sub-scales as in the draft scale.

Reliability:
t

The Tamily Climate Scale’ was administered to a sample o f 90


adolescent subjects, both boys and girls, and 87 completed forms were
obtained. ITie scores for odd and even items in the questionnaire were
scorcd separately for each variable, and the split-half reliability coefficients
were calculated for each variable, after correction using Spearman-Brown
prophecy formula (Garrett, 1969). 'i'he split-half reliability coefficients
obtained for each variable is given below:

1. Paternal Control : 0. 84

2. Maternal Control : 0. 81

3. Paternal Affect : 0. 73
4. Maternal Affect : 0. 79

5. Parental Harmony : 0. 82

6. Sibling Relationship : 0. 76 (all significant at 0.01 level).

I'hese results indicate that this is a reliable instrument for measuring


the family climate variables.

Validity:

As most o f the items were eitlier adapted from well-known and


standardized instruments having established reliability and validity or were
evaluated by a number o f experts, the scale may be said to possess content
validity.

b. Mental Health Profile:

Scoring was done as in Family Climate Scale. Item analysis for the
four sub-scales (Self-Esteem, Autonomy, Emotional Stability, and Social
Competence) were done separately,
r

1. Out o f the 15 items in the Self-esteem draft scale, 13 had t-values


significant at 0.01 level. Out o f these 10 were selected , on the
basis o f the three criteria mentioned above, for the final test.

2. Among 15 items in the Autonomy scale, 12 items had t-values


significant at 0.01 level, out o f which 10 were selected for the

final test.
3. All the 15 items in the emotional stability scale had t-values
significant at 0.01 level, from which 10 were included in the final
test.

4. In the case o f Social competence, 13 out o f the 15 items had t-


vales significant at 0.01 level and 10 were included in the final
test.

Reliability:

Reliability was estimated separately for each o f the four sub-scalcs.


Mental Health Profile was administered along with Family Climate Scale to
90 subjects, and the 87 completed forms were scored and subjected to split-
half reliability. Odd and even items were scored separately and the
reliability coefficients were calculated after correction using Spearman-
Brown prophecy formula. The reliability coefficient obtained for the four
sub-scalcs were:

1. Self-Esteem 0.91

2. Autonomy : 0. 77

3. Emotional Stability : 0. 76

4. Social Competence ; 0. 81

All are significant at 0.01 level. This


reliable one.
Validity:

Most o f the items were either adapted from existing scales or


selected after consulting a number o f experts. So these scales may be said to
have content validity. ,

5. PROCEDURE USED FOR DATA COLLECTION

As discussed earlier in this chapter, the subjects were selected from a


wide variety o f istitutions. With prior permissions from the authorities
concerned, the subjects were met in small groups o f ten to fifteen students
and were briefed about the purpose o f the research and also assured o f the
confidentiality o f the data provided.

Either English or Malayalam versions o f the tests were handed out


to the subjects according to their preference. The instructions were read out
to the subjects by the researcher and clarifications were made. I'hc sujects
were allowed to fill in the form and doubts, if any, were cleared in the
process. Since the subjects were met in small groups, they were not allowed
to consult one another while responding to the statements. The subjects
were prompted to give genuine responses for the successful completion o f

the research.

About 35 to 45 minutes were taken by a subject to complete the test.


Out o f a total o f 629 data sheets, 592 were complete in all respects, and were

used for final analysis;


6.THE STATISTICAL TECHNIQUES

The following statistical techniques were used for the analysis o f the
data, with the help o f a computer:

1. Pearson’s product-moment coefficients o f correlations among


all the ten variables (Garrett, 1969).

2. Principal axes method o f factor analysis for the eight variables


(Fruchter, 1954).

3. The t-test for large independent samples for the ten variables
between boys and girls, between urban and rural subjects, between
urban boys and urban girls and between rural boys and rural girls
(Garrett, 1969).

4. Analysis o f variance among three age groups - early, middle,


and late adolescents (Garrett, 1969).

5. Scheffe procedure (Schcffc, 1964).

Interpfetation of Correlations:

Garrett (1969) provides the following classification for interpreting


the various values o f r, which is given below

relationship
r from ± 0.20 to + 0.40 denotes low correlation;

r from ± 0.40 to ± 0.70 denotes substantial or

marked relationship; and

r from ± 0.70 to ± 1.00 denotes high to very high

relationship.

Garrett (1969) also notes that this classijScation can be accepted


only as general guideline with certain reservations. The coefficient of
correlation must always be judged with regard to :

1.The nature o f variable with which we are dealing;

2.'I'he significance o f the coefficient;

3.11ie variablity o f the group;

4.The reliability coefficients o f the tests used; and

5. The purpose for which the r was computed.


r‘

In the present study this general scheme is accepted as such.

The statistical significance o f coefficient o f correlation has to be


considered before the correlation itself is interpreted. The r may be tested
I

agjiinst the hypothesis that the population r is zero. If the r for the sample
is large enough to invalidate this null hypothesis, we accept r as indicating
the presence o f at least some degree o f correlation. To test this, the
obtained r is compared with the limits established using the standard error

o f r (Garrett, 1969). The standard error for the null hypothesis is 1 /V n .

ITence, a coefficient o f correlation will be significant at 0.01 level if it

exceeds I /V n x 2.58, and at 0.05 level, if it exceeds 1 /V n x 1.96.

Factor Analysis:

Factor analysis as a powerful multivariate tool has been used to test


the second hypothesis for the present study. According to Fruchter (1954) :

Factor analysis starts with a set o f obsevations


obtained from a given sample ... It is a method o f analysing
this set o f obsevations from their intercorrelations to detrmine
whedier the variations represented can be accounted for
adequately by a number o f basic categories smaller than that
with which the investigation was started. ITie data obtained
with a large number o f a priori measures may be explained in
terms o f a smaller number o f reference variables.

According to I larman (1960) :


j

The principal concern o f factor analysis is the


resolution o f a set o f variableslinearly in terms o f (usually) a
smaller number o f categories or ‘factors’. This resolution can
be accomplished by the analysis o f the correlations among the
variables. A satisfactory solution will yield factors which
convey all the essential information o f the original set o f
variables. Thus, the chief aim is, to attain scientific parsimony
or economy o f description.

A given matrix o f correlation can be factored in several ways. The


present study used the principal axes method, which is one o f the more
Gt s m ^

popular procedures o f factor analysis. The advantages o f this method are


that:

liach factor extracts the maximum amount o f variance


(i.e., the sum o f squares o f the factor loadings is maximised on
cach factor) and g^ves the smallest possible residuals. The
correlation matrix is condensed into ^ e smallest number o f
orthogonal factors by this method. The method also has the
advantage o f giving mathematically unique (least square)
solution for a given table o f correlations (Fruchter, 1954).

Te main limitation o f this method is that it involves considerable


amount o f computations. But this has been overcome by the use o f
computer facilities.

Interpretation of factors:

I'hc following conventions are used for interpreting the


factors (Fruchter, 1954) :

1. Factor loadings o f ± 0.20 or less insignificant


1
2. i'actor loadings o f ± 0.20 to ± 0.30 - low

3. Factor loadings o f ± 0.30 to ± 0.50 - moderate


4

4. Factor loadings o f ± 0.50 to ± 0. 70 - high

5. Factor loadings above ± 0.70 - very high.


The t-Test:

I ’hc t-test (Garrett, 1969) was used to compare adolescent groups,


based on gender, residence (urban-rural), and medium o f instruction, on
their perceptions o f family climate and their mental health.

The t-test for the large independent samples have been applied
(two-tailed tests) as the sample size in each o f the categories exceeded 30.

When t-value exceeds a cut off point (depending on degree o f


freedom) the difference in means is considered significant, and the null
hypothesis is rejected. When the t-value is below the cut o ff point, the
difference is said to be not significant, and the null hypothesis is accepted.

For the present study the df = (N1 + N 2 —2) is 590. Hence the
cut o ff point is taken as 1. 96 at 0 . 05 level.

T h e Analysis o f Variance: ^
I

The comparison o f difference between me^ns for two groups is


•i
usually done using the t-test, as the procedure is simple. But, when three or
more groups arc involved in the comparison, the standard error for the t-
test becomes complicated. In such cases, another procedure, using the F
distribution, known as analysis of variance (ANOVA), is commonly used. It
requires the same assumptions as the t-test: independent and random
samples, normal distribution o f the dependent variables in each population
group.
The F-ratio is calculated as per the procedure explained in Garrett
(1969). In the present study there are 592 subjects belonging to three
different age groups. Therefore, there are 589 and 2 degrees o f freedom.
The cut o ff points were found out from the Tables (Garrett, 1969), the
value being 3.01 at 0.05 level.

The Scheffe Procedure:

When a significant F-ratio is obtained, it is left o f with the


investigator to go for further analysis in order to identify the groups which
indicate significant differences, when there are more than two groups
involved. For the purpose o f the present study Scheffe procedure (Scheffe,
1964) is followed. As this is a powerful test, the level o f significance is fixed
at 0.05 level.

The detailed analysis and interpretation o f the data are given in the
next chaptcr.

You might also like