Professional Documents
Culture Documents
research-article2014
Article
Cross-Cultural Research
2015, Vol. 49(1) 90105
2014 SAGE Publications
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/1069397114548647
ccr.sagepub.com
Abstract
This study used multiple-group confirmatory factor analysis procedure to
examine the measurement and structural invariance across Malaysian Malay
(n = 724), Chinese (n = 372), and Indian (n = 259) parent ratings of their
children on the short version of the ParentParental AcceptanceRejection
Questionnaire (Parent-PARQ/S). The Parent-PARQ/S measures parents
perceptions of their rejection of their children and has scales for coldness/lack
of affection, hostility/aggression, indifference/neglect, and undifferentiated
rejection. Results provided strong support for the measurement and
structural invariance of the Parent-PARQ/S. In addition, the three ethnic
groups did not differ for all four latent mean scores.
Keywords
Parental AcceptanceRejection Questionnaire, measurement and structural
invariance, invariance for ethnic groups, Malaysia
1University
2University
of Ballarat, Australia
of Tasmania, Hobart, Australia
Corresponding Author:
Rapson Gomez, School of Health Science, University of Ballarat, University Drive, Mt Helen,
P.O. Box 663, Ballarat, Victoria 3353, Australia.
Email: rapson.gomez@ballarat.edu.au
91
92
or fathers treatment of them when they were about 7 through 12 years old.
Meta-analysis of this measure for its internal consistency across many countries has shown homogeneity in the effect size (Khaleque & Rohner, 2002a,
2002b), and factor analytic studies have yielded the same factor structure
across many countries (e.g., Comunian & Gielen, 2001; Rohner & Cournoyer,
1994). These findings raise the possibility of, but do not guarantee, measurement equivalency or measurement invariance of the PARQ across different
cultural groups. For a rating scale, measurement invariance deals with
whether the items in the rating scale have the same scale properties when
completed by individuals from different groups, such as boys and girls
(Brown, 2006). If there is weak or no support for invariance, then it follows
that the individuals from the different groups examined cannot be justifiably
compared on the raw scores as the raw scores are confounded by differences
in measurement and scaling properties that are group specific. The opposite
is the case when there is support for measurement invariance.
A powerful method for testing invariance is multiple-group CFA. This
procedure can test invariance for the measurement and structural components
of a model. For the measurement model, it can test for configural invariance
(equality for form across groups), metric invariance (equality for factor loadings across groups), thresholds invariance (equality for responses to response
categories across groups), and error variances invariance (equality for uniqueness across groups). For the structural model, it can test for invariance for
factor variances and covariances across groups, and compare the groups for
latent mean scores. To date, only one study has examined invariance across
national groups for the PARQ (Gomez & Rohner, 2011). The study found
support for all types of measurement and structural invariance across ratings
of the Adult-PARQ provided by Australian and U.S. adults. Despite this and
in view of the limited cross-cultural data for the PARQ, it seems prudent to
conduct more studies in this area, involving other versions of the PARQ and
for other cultures.
Among the group of PARQ measures is the short version of the ParentPARQ or Parent-PARQ/S (Rohner, 2005). As noted previously, this measure
asks parents to assess the way they now treat their children, and it provides
scales for measuring coldness/lack of affection, hostility/aggression, indifference/neglect, and undifferentiated rejection. A recent study examined the factor structure, reliability, and validity of this measure (Gomez & Suhaimi,
2014). It also examined its measurement invariance for parent ratings of boys
and girls, and mothers and fathers ratings of their children. Findings based
on CFA indicated only marginal fit for the original four-factor model. There
was excellent fit for an alternate four-factor model in which Item 13, which
loads in the original model on the indifference/neglect factor, was loaded on
93
Item # 1
Item # 3
Item # 9
Item # 12
Item # 13
Item # 17
Coldness/
lack of affection
Item # 19
Item # 22
Item # 24
Item # 2
Item # 7
Item # 11
Hostility/
Aggression
Item # 15
Item # 23
Item # 4
Item # 6
Item # 10
Item # 14
Indifference/
Neglect
Item # 18
Item # 20
Item # 5
Item # 8
Item # 16
Undifferentiated
rejection
Item # 21
Figure 1. Path diagram of the adjusted four-factor PARQ model used in the
multiple-group CFA analyses.
Note. For clarity, error variances are not shown. Error variances were not correlated. PARQ
= Parental AcceptanceRejection Questionnaire; CFA = confirmatory factor analysis.
the coldness/lack of affection factor instead. Figure 1 shows the path diagram
of the revised model (henceforth referred to as the adjusted four-factor
model). The adjusted model showed acceptable internal consistencies
(Cronbachs coefficients) for all factors (ranging from .77 to .65), with all
94
95
were similar for ambition, filial piety, honesty, knowledge, and trustworthiness, whereas the Malays and Indians were similar for piousness. The Chinese
and Indians were similar for rituals and traditions. Overall, therefore, based
on Hofstedes definition of culture, Malaysian Malays, Chinese, and Indians
can be viewed as culturally distinct groups.
The different cultural values of Malays, Chinese, and Indians can be
expected to influence their attitudes and behaviors differently, including parenting style and interactions with their children (Keshavarz & Baharudin,
2009). Consistent with this, Ang (2006) found different effects for parenting
styles on Malay and Chinese adolescents personal and social behaviors. Bao
and Xu (2006) found that Malaysian Malay adolescents were more attached
to their parents and peers, compared with Malaysian Chinese adolescents.
Quah (2004) examined differences in parenting styles and expectations
across Malay, Chinese, and Indian parents in Singapore (Malaysias neighbor). She found that in comparison with Indian and Malay parents, Chinese
parents were more likely to consider physical punishment (in contrast to reasoning and rules) as being effective for disciplining children and were less
likely to demonstrate their affection by hugging, kissing, and holding their
children. Between Indian and Malay parents, Indian parents were less indulgent, stricter, and had higher expectation of behavior discipline and standards
from their children. Consistent with the findings for Chinese parents, Chi,
Baharudin, & Hossain, 2012) showed that Malaysian Chinese parents use
relatively more authoritative parenting styles (characterized by high control
over children and emphasizing conformity, discipline, and obedience from
children) than authoritarian (parenting that display democratic and negotiated
interactions and warmth, and encouragement of independence in children)
and permissive (non-demanding, non-controlling) parenting styles. Indeed,
Confucianism, which is central to Chinese values, requires Chinese children
to be obedient, self-disciplined, hard-working, and achievement-oriented,
and for Chinese parents to train (guan) their children directly through high
levels of control and very close monitoring. Given such cultural differences
(including ethnic and parenting differences), it is conceivable that they would
differentially affect parental perceptions and acceptability of their childrens
behaviors. In brief, Chinese parents can be expected to require better child
behavior standards and are likely to be less tolerant of their childrens misbehaviors than Indian and Malay parents. Also, Indian parents can be
expected to require better child behavior standards and are likely to be less
tolerant of their childrens misbehaviors than Malay parents. Given these
differences, it is conceivable that Chinese, Malay, and Indian parents could
have different interpretations of comparable parenting behaviors that could
differentially influence their responses on the Parent-PARQ. This could, in
turn, contribute to lack of measurement invariance across their ratings.
Downloaded from ccr.sagepub.com at Universiti Teknologi Malaysia on April 7, 2016
96
The aim of this study was to use multiple-group CFA to test measurement
and structural invariance across Malaysian Malay, Chinese, and Indian parent
ratings of their children for the Parent-PARQ/S. As our previous study involving the same sample as in this study showed strong support for the adjusted
four-factor model, we used this model in all our analyses. It was speculated
that as there is evidence of parenting differences across Malay, Chinese, and
Indian parents, there will be some degree of non-invariance across their
ratings.
Method
Participants
The participants were 1,382 caregivers (primarily parents, and henceforth
referred to as parents) of primary schoolchildren from 12 schools in Klang
Valley of the State of Selangor in Malaysia. This state was selected as it is
highly representative of the ethnic distribution of West Malaysia. The respondents provided ratings for 603 (43.6%) boys and 779 (56.4%) girls, aged
between 5 and 13 years. The mean age (SD) for boys was 9.79 years (1.78),
and it was 9.92 years (1.81) for girls. There was no significant difference for
age between boys and girls, t(1,380) = 1.25, ns. In terms of ethnicity, 729
(52.4%) were Malays, 372 (26.9%) were Chinese, 259 (18.7%) were Indians,
and 27 (2.0%) belonged to the other categories. For Malaysia as a whole, the
figures are 61.5% Malays, 26.4% Chinese, and 10.5% Indian (Department of
Statistics Malaysia, 2010). The distribution of the ethnic groups in the study
sample did not differ significantly from the Malaysian population, 2(2) =
3.03, ns. In terms of educational attainment, most parents completed primary
and secondary education, and the majority of fathers were in skilled/semiskilled (e.g., technicians and associated technicians, 25%) and service-related
(e.g., service workers in shops and markets; 40%) occupations. Professionals
(e.g., teachers, nurses, doctors, etc.) constituted around 10%. Mothers completed 1,101 (79.7%) of all the parental ratings. Father completed 222 (16.1%)
ratings.
Measure
The measure of relevance in this article is the Parent-PARQ/S. The ParentPARQ/S, completed by parents, has 24 items that measure parental acceptance and rejection of their children and adolescents. It has four scales:
coldness/lack of affection (eight items), hostility/aggression (six items),
indifference/neglect (six items), and undifferentiated rejection (four items).
97
Procedure
Prior to data collection, ethical approval for the study was obtained from the
University of Tasmania Social Sciences Human Research Ethics Committee.
Following this, approvals were sought from the relevant authorities for distributing the relevant documents, including the dissemination of the ParentPARQ/S to parents through schools. Initially, approval was obtained from the
Research Promotion and Coordination Committee, Economic Planning Unit
(EPU), Prime Ministers Department, Malaysia, to conduct the research
through the primary schools in Klang Valley in the State of Selangor in
98
99
Statistical Procedures
All the CFA analyses in the study were conducted using Mplus (Version 6.1)
software (Muthen & Muthen, 2010), and they used the mean and variance
adjusted weighted least squares (WLSMV). This is a robust estimator for
ordered-categorical scores. According to Brown (2006), the WLSMV estimator provides the best option for modeling categorical data.
Multiple-group CFA measurement invariance was tested using the procedure proposed by Millsap and Yun-Tein (2004) for the WLSMV estimator
with theta parameterization. In brief, this procedure involves comparing progressively more constrained models that test for measurement invariance:
configural invariance (equality for form), metric invariance (equality for factor loadings), thresholds invariance (equality for responses to response categories), and error variances invariance (equality for uniqueness). According
to Millsap and Yun-Tein, for baseline model identification, the following
restrictions are necessary: (a) The factor loadings of the scaling indicators
must be set to 1 for all group; (b) the thresholds of these scaling
indicators must be constrained to be equal across all groups; (c) for all
other indicators, their first threshold and additional thresholds (if needed)
need to be constrained equal across groups; and (d) residual variances need to
be constrained to 1 in the first group and to be freely estimated in the other
groups. When there is some support for measurement invariance (full or partial), structural invariance (equivalency for factor variances and covariances)
can be examined. Also, when there is some support for measurement invariance, the groups can be compared for latent mean scores, taking into account
the non-invariance in the measurement model.
The goodness-of-fit of the CFA models was examined using WLSMV2.
Like all other 2 values, WLSMV2 values are also inflated by large sample
sizes. In addition to the WLSMV2, Mplus also provides approximate (or
practical) fit indices. These are the root mean squared error of approximation
(RMSEA), the comparative fit index (CFI), and the TuckerLewis Index
(TLI). These approximate fit indices were also used in this study to evaluate
the goodness-of-fit of models. The guidelines suggested by Hu and Bentler
(1998) are that RMSEA values close to 0.06 or below be taken as good fit,
0.07 to <0.08 as moderate fit, 0.08 to 0.10 as marginal fit, and >0.10 as poor
fit. For the CFI and TLI, values of 0.95 or above are taken as indicating good
model-data fit, and values of 0.90 and <0.95 are taken as marginally acceptable fit. The various CFA invariance models were compared using the
100
WLSMV2 difference test that was computed using the option available in
Mplus. An value of .01 was used to allow for more stringent Type 2 error
control in models compared.
Results
Single Group CFA of the Adjusted Four-Factor Model for the
Different Ethnic Groups
Initially, the fit values of the adjusted four-factor model were examined separately for ratings provided by Malay (n = 724), Chinese (n = 372), and Indian
(n = 259) parents. The fit indices of Malay ratings were 2(df = 246) = 536.24,
p < .001, RMSEA = 0.040, CFI = 0.969, TLI = 0.965. The fit indices of
Chinese ratings were 2(df = 246) = 395.63, p < .001, RMSEA = 0.040, CFI
= 0.971, TLI = 0.967; and the fit indices of Indian ratings were 2(df = 246) =
392.76, p < .001, RMSEA = 0.048, CFI = 0.959, TLI = 0.954. The RMSEA,
CFI, and TLI values for all three cultural groups indicated good fit, thereby
supporting the construct validity of the adjusted four-factor model of the
Parent-PARQ/S for all three ethnic groups.
101
Table 1. Results of Tests for Invariance Across PARQ Ratings of Malay, Chinese,
and Indian Parents.
Model fit
Models (M)
M1: Configural invariance
M2: Metric invariance
M3: Thresholds invariance
M4: Error variances invariance
M4A: M4 with error variances
for Item 16 free
M5: Invariance for the all latent
factor variances
M6: Invariance for the
covariances between latent
factors
M7: Invariance for the means of
the latent factors
Model difference
df
RMSEA
CFI
TLI
df
1363.87
1369.00
1375.37
1431.12
1405.71
788
828
866
914
913
0.040
0.038
0.036
0.035
0.035
0.967
0.969
0.971
0.970
0.972
0.965
0.969
0.972
0.973
0.974
M2 M1
M3 M2
M4 M2
M4A M2
40
38
86
85
48.57
37.42
135.84***
113.48
1356.69
874
0.035
0.972
0.974
M5 M3
8.74
1307.85
878
0.033
0.975
0.977
M6 M3
12
7.93
1378.71
874
0.036
0.971
0.973
M7 M3
14.40
Note. All WLSMV2 values were significant (p < .001). PARQ = Parental AcceptanceRejection
Questionnaire; 2 = weighted least square with mean and varianceadjusted chi square (WLSMV2);
RMSEA = root mean square error of approximation; CFI = comparative fit index; TLI = TuckerLewis
Index.
***p < .001.
rejection factor. For this item, Indians had lower error variance than the
Malay and Chinese groups (completely standardized values: Malay = .720,
Chinese = .705, Indian = .587). Taken together, however, these results provide quite good support for the measurement invariance of the PARQ across
the ratings provided by Malay, Chinese, and Indian parents.
As also shown in Table 1, for the structural model, there was support for
invariance for all factor variances and covariances (M5 and M6, respectively,
in Table 1) as these models did not differ significantly from the thresholds
invariance model (M3 in Table 1), WLSMV2(df = 8) = 8.74, ns; and
WLSMV2(df = 12) = 7.93, ns, respectively. As shown in Table 1, the analysis showed support for invariance for the factor mean scores model (M7 in
Table 1), as this model did not differ from the thresholds invariance model
(M3 in Table 1), WLSMV2(df = 8) = 14.40, ns.
Discussion
The aim of the current study was to use multiple-group CFA analyses to
examine measurement and structural invariance for Malaysian Malay,
Chinese, and Indian parent ratings of their children on Parent-PARQ/S. For
the measurement model, the findings showed support for the configural
102
model (same pattern of factor structure), and for full measurement invariance
for the metric (same factor loadings), and thresholds (same response categories) models. Of the 24 items in this questionnaire, 23 items showed invariance for error variances (same unique variances). The item that did not show
invariance was Item 16 (I make my child feel unloved if he/she misbehaves), which is part of the undifferentiated rejection factor. For this item,
Indians had lower error variance than the Malay and Chinese groups.
Although the test for error variances is often included during invariance testing, most methodologists consider this test as overly stringent and unnecessary (Brown, 2006). Taken together, therefore, the results can be taken as
providing very good support for the measurement invariance of the ParentPARQ/S across the ratings provided by Malaysian Malay, Chinese, and
Indian parents. This means that the ratings provided by these groups of parents have the same measurement and scaling properties, and therefore, their
observed scores can be compared directly.
For the structural model, the findings showed invariance for all factor
variances and covariances. This means that the scores for the four latent factors (hostility/aggression, indifference/neglect, and indifference/neglect) in
the adjusted four-factor model do not differ for ratings provided by Malaysian
Malay, Chinese, and Indian parents in terms of their variability and intercorrelations, respectively. The findings for the structural model also showed no
difference between Malay, Chinese, and Indian parent ratings for the latent
scores for coldness/lack of affection, hostility/aggression, indifference/
neglect, and indifference/neglect. All the invariance findings suggest that the
cultural differences of Malays, Chinese, and Indians that influence the way
parents treat and interact with their children (Ang, 2006; Bao & Xu, 2006;
Quah, 2004) appear not to differentially influence the way they rate their
children on the Parent-PARQ/S. However, this does not imply that cultural
factors are irrelevant for understanding parenting behaviors and attitudes. It
may be that such factors are relevant for understanding other issues, such as
the processes in play that lead to similar parenting behaviors and attitudes
across different cultural groups.
At a general level, the findings support the use of the Parent-PARQ/S for
cross-cultural studies of parental rejection (Rohner, 2005). However,
because this study only involved three cultural groups in one nation, more
studies are needed to establish whether this can be generalized to cultures
in other countries. Furthermore, although the findings in this study have
been interpreted in terms of cross-cultural invariance for the ParentPARQ/S, this interpretation is based on the assumption that the Malay,
Chinese, and Indian communities in Malaysia are culturally distinct.
However, if this assumption is incorrect to start with, then it is
103
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
104
References
Ang, P. R. (2006). Effects of parenting style on personal and social variables for
Asian adolescents. American Journal of Orthopsychiatry, 76, 503-511.
Bao, K. B., & Xu, Q. M. (2006). A comparison of attachment in adolescents in mainland China and Malaysia. Chinese Journal of Clinical Psychology, 14, 172-174.
Brown, T. A. (2006). Confirmatory factor analysis for applied research. New York,
NY: The Guilford Press.
Calmorin, L. P., & Calmorin, M. A. (Ed.). (2007). Research methods and thesis writing (2nd ed.). Manila, Philippines: Rex Book Store.
Chi, Y. H., Baharudin, R., & Hossain, Z. (2012). Fathers parenting styles in Chinese
families in urban Malaysia. Pertanika Journal of Social Science & Humanities,
20, 1019-1034.
Comunian, A. L., & Gielen, U. P. (2001). An Italian study of parental acceptancerejection. In R. Roth & S. Neil (Eds.), A matter of life: Psychological theory,
research, and practice (pp. 166-175). Lengerich, Germany: Pabst Science
Publishers.
Dahlan, H. M. (1991). Local values in intercultural management. Malaysian
Management Review, 1, 45-50.
Department of Statistics Malaysia. (2010). Vital statistics: Malaysia 2010. Putrajaya:
Author.
Gomez, R., & Rohner, R. P. (2011). Tests of factor structure and invariance in the
US and Australia using the adult version of the Parental AcceptanceRejection
Questionnaire. Cross-Cultural Research, 45, 267-285.
Gomez, R., & Suhaimi, A. F. (2014). Malaysian parent ratings of the short version
of the Parent Parental Acceptance-Rejection Questionnaire: Factor structure and
invariance across sex, and mother and father. Parenting: Science and Practice,
14, 195-214.
Hofstede, G. (1980). Cultures consequences: International differences in workrelated values. Beverly Hills, CA: Sage.
Hu, L. T., & Bentler, P. M. (1998). Fit indices in covariance structure modeling:
Sensitivity to underparameterized model misspecification. Psychological
Methods, 3, 424-453.
Idris, A. (2011). Ethnicity and cultural values: Am empirical study of Malay and
Chinese entrepreneurs in Peninsular Malaysia. Sarjana, 26, 22-40.
Keshavarz, S., & Baharudin, R. (2009). Parenting style in a collectivist culture of
Malaysia. European Journal of Social Sciences, 10, 66-73.
Khaleque, A., & Rohner, R. P. (2002a). Perceived parental acceptancerejection and
psychological adjustment: A meta-analysis of cross-cultural and intracultural
studies. Journal of Marriage and Family, 64, 54-64.
Khaleque, A., & Rohner, R. P. (2002b). Reliability of measures assessing the relation
between perceived parental acceptancerejection and psychological adjustment:
Meta-analysis of cross-cultural and intracultural studies. Journal of CrossCultural Psychology, 33, 87-99.
105
Author Biographies
Rapson Gomez is professor of clinical psychology at Federation University, Ballarat,
Australia. His major research interests are on child and adolescent psychopathology,
and personality.
Aida Farhana Suhaimi is a clinical psychologist, currently practicing in Malaysia.