Page 1
JNM
Journal of Nursing Measurement
l f
i
www.springerpub.com/jnm
With the Compliments of Springer Publishing Company, LLC
Page 2
Journal of Nursing Measurement, Volume 19, Number 3, 2011
© 2011 Springer Publishing Company
http://dx.doi.org/10.1891/10613749.19.3.131
131
14Item Resilience Scale (RS14):
Psychometric Properties of the
Brazilian Version
Bruno Figueiredo Damásio, MPsy, PhDc
Juliane Callegaro Borsa, MPsy, PhDc
Universidade Federal do Rio Grande do Sul, Porto Alegre
Joilson Pereira da Silva, PhD
Universidade Federal de Sergipe, Itabaiana
The Resilience Scale (RS) was developed to evaluate the levels of resilience in the general
population. Its reduced version (RS14) has presented reliable internal consistency and
external validity. However, its psychometric properties have not been systematically
evaluated. The objective of this study was to present the psychometric properties of the
Brazilian RS14. A total of 1,139 subjects selected by convenience (62.9% women) from
14 to 59 years old (M 5 26.1, SD 5 11.61) participated in the study. Exploratory factor
analyses (EFAs) and parallel analysis were conducted in order to assess the factor structure
of the scale. A 13item singlefactor solution was achieved. Confirmatory factor analyses
(CFA) and multigroup CFA (MGCFA) corroborated the goodness of fit and measurement
invariance of the obtained exploratory solution. The levels of resilience correlated
negatively with depression and positively with meaning in life and selfefficacy.
Keywords: resilience; scale; validation; multigroup confirmatory factor analysis; parallel
analysis
I
2010). As a key factor in the process of overcoming and adapting from negative events,
resilience has become one of the most researched topics in the mental health field (Oshio,
Kaneko, Nagamine, & Nakaya, 2003).
The theoretical and empirical literature on resilience reflects little consensus about
its definition, with substantial variations in operationalization and measurement of key
constructs (Luthar, Cicchetti, & Becker, 2000). Some studies comprehend resilience as
an individual characteristic that moderates the negative effects of stress and promotes
positive adaptation (Charney, 2004; Wagnild & Young, 1993), whereas others explicitly
rejected this individual trait model (Masten, 2001; Masten, Best, & Garmezy, 1990). This
latter perspective understands resilience as a dynamic process involving personal strengths
and capacities, external resources such as a healthy family environment, and presence of
external support systems that reinforce efficient coping and adaptive adjustment. In others
n the last decades, there has been an increasing interest in understanding the strengths that
are associated with healthy adjustment trajectories, such as resilience (CampbellSills &
Stein, 2007; Ryan & Caltabiano, 2009; von Soest, Mossige, Stefansen, & Hjemdal,
Copyright © Springer Publishing Company, LLC
Page 3
132 Damásio et al.
words, resilience is conceptualized as a multidimensional construct that varies with con
text, time, age, gender, social support systems, family environments, and cultural origins,
as well as individual capabilities (Luthar et al., 2000; Masten & Wright, 2010).
Because of the growing interest in this concept and the necessity of briefly evaluating the
ability to cope with different situations, some scales have been developed to measure the
resilience construct. Two examples of instruments for assessing resilience in children and
adolescents are the Adolescent Resilience Scale (Oshio et al., 2003) and the Resilience Scale
for Adolescents (READ; von Soest et al., 2010). Addressed to assessing adults resilience, a
greater number of instruments are available, as for example, the Resilience Scale for Adults
(RSA; Friborg, Barlaug, Martinussen, Rosenvinge, & Hjemdal, 2005), the Resilience in
Midlife Scale (RIM Scale; Ryan & Caltabiano, 2009), the ConnorDavidson Resilience
Scale (CDRISC; Connor & Davidson, 2003), the Brief Resilience Scale (BRS; Smith et al.,
2008), and the Brief Resilient Coping Scale (BRCS; Sinclair & Wallston, 2004).
One of the most widely used instruments in resilience research is the Resilience Scale (RS)
by Wagnild and Young (1993). The RS is the first instrument developed to measure resilience
and can be applied in a wide variety of age groups, from adolescents to older people (Ahern,
Kiehl, Sole, & Byers, 2006). The RS was developed from a qualitative study of 24 women who
showed positive psychosocial adaptation from different life events (Wagnild & Young, 1990).
Each woman was asked to tell how they conducted a selfidentified loss. According to their
narratives and followed by a review of the literature, the authors acknowledged five common
components identified to be personal constituents of resilience: equanimity, perseverance, self
reliance, meaningfulness, and existential aloneness (Wagnild, 2010; Wagnild & Young, 1993).
Equanimity is referred as a balanced perspective of life and experiences and might be
viewed as sitting loose and taking what comes, thus moderating the extreme responses to
adversity, a construct often related to the sense of humor. Perseverance is the ability to
keep going despite setbacks, generally found in people who tend to recognize and rely on
their personal strengths and capabilities. Selfreliance is considered a selfefficacy belief
specially linked to problemsolving skills. In general, this ability is achieved with life
experiences and is most frequently encountered in people who comprehends and accepts
their own capabilities and limitations. Meaningfulness is the belief that life has a purpose
and recognition that there is a reason for which to live. Finally, existential aloneness is the
realization that each person is unique and that although some experiences can be shared,
others must be faced alone (Wagnild, 2009; Wagnild & Young, 1993).
The first version of the Resilience Scale (RS25; Wagnild & Young, 1993) consisted in a
25item instrument aiming to evaluate the individual resilience degree through the five personal
characteristics aforementioned. Used in a random sample of 810 North American older adults,
the authors found, through principal component analysis (PCA) and oblimin rotation, a
twofactor solution as the most reliable. The first factor, titled “Personal Competence,” was
composed of 17 items. The second factor, titled “Acceptance of Self and Life” was constituted
by eight items. The twofactor solution explained 44% of the construct variance.
After its original publication, the RS25 has been translated and adapted for many
languages in several countries such as Brazil (Pesce et al., 2005), Argentina (Rodríguez
et al., 2009), Sweden (Nygren, Randström, Lejonklou, & Lundman, 2004), Japan
(Nishi, Uehara, Kondo, & Matsuoka, 2010), and Spain (Heilemann, Lee, & Kury, 2003).
The RS25 has been consistently reliable with alpha coefficients ranging from .84 to .94.
Convergent validity as well as testretest reliability has also been extensively presented in
these validation and adaptation studies.
Although many validation criteria have been consistently reported, the RS25 factor
structure has not been stable and well clear, suggesting the need for further analyses,
Copyright © Springer Publishing Company, LLC
Page 4
Psychometric Properties of Brazilian RS14 133
as predicted in the original study (Wagnild & Young, 1993). Refinement studies led to
the construction of a shorter instrument (RS14), consisted by 14 items from the original
RS25 (Wagnild, 2010). The nine items excluded were those that showed an interitem
correlation above .40 in the author’s previous studies (Wagnild, 2010).
The validation study of the RS14 was conducted in a sample of 690 middleaged
and older adults. The 14 items from the RS14 were entered into a PCA using oblimin
rotation. A onefactor solution accounting for 53% of the total variance was found. All
items loaded higher than .40, ranging from .42 to .64. Cronbach’s alpha reliability was
.93. The RS25 and RS14 were strongly correlated (r 5 .97, p , .001). Additional
recent analysis of the RS14 using the same extraction and rotation methods (PCA and
oblimin, respectively) in another sample of 1,161 individuals (average 36.4 years old)
strongly supported the onefactor solution (Cronbach’s a 5 .91, explained variance 5
46%; Wagnild, 2010).
Searching in the literature, only one study reporting the psychometric properties of the
RS14 was found (Nishi et al., 2010). Conducting a PCA, as used in the original study,
Nishi et al. (2010) also found a singlefactor solution. All items loaded onto the first
component and had factor loadings greater than .49. The total variance was 39.4%. In this
study, the RS14 showed a negative correlation with depression symptoms (r 5 2.28,
p , .01) and social disability (r 5 2.32, p , .01) and positive correlations with self
esteem (r 5 .28, p , .01) and social support (r 5 .38, p , .01).
By now, the RS14 has shown a reliable convergent validity as well as an invariant
factor structure (Nishi et al., 2010; Wagnild, 2010). However, no confirmatory factor
analysis (CFA) study has been reported testing the goodness of fit of the RS14. Because
of its theoretical relevance and the need for a brief scale to evaluate the degree of people’s
resilience in the general population in Brazil, this study aimed to evaluate the reliability
and validity of the Brazilian RS14.
METHOD
Participants
The sample consisted of 1,139 subjects (62.9% women), from 14 to 59 years old (M
5 26.1, SD 5 11.61), residing at northeast of Brazil. Data were obtained from two
independent studies. The first sample was composed of 629 youngsters (252 males and
377 females; ranging in age from 14 to 29 years old, with a mean age of 17.4 years old;
SD 5 2.44), recruited to participate in this study as part of a larger research project
focused on psychosocial aspects related to young resilience levels. The second sample was
composed of 510 school teachers (171 males and 339 females; ranging in age from 18 to
59 years old, with mean age of 36.21 years old; SD 5 9.93), recruited to participate in a
previous study, which aimed to evaluate the levels of psychological wellbeing and per
sonal and contextual characteristics related to the process of teacher’s resilience. For the
objective of this study, both samples were grouped together.
Procedures
For the first sample, participants were asked to respond to the instruments at different
scholarly contexts (schools and preuniversity courses), from 13 public and private
institutions. For the second sample, participants were asked to respond to the instruments
in their workplaces. Teachers from 57 public and private schools were assessed. Both data
Copyright © Springer Publishing Company, LLC
Page 5
134 Damásio et al.
collections occurred after previous authorizations of the State University of Paraíba Ethics
Committee (Institutional Review Board). Participants provided written informed consent
before the data were collected.
Measures
14Item Resilience Scale (RS14). The Brazilian version of the RS14 used in this
study was based on the adaptation process developed by Pesce et al. (2005), in which
the RS25 was translated and validated to the Brazilian sample. For the objective of this
study, the nine items of the Brazilian RS25 that did not compose the reduced version were
deleted. The current version of the RS14 contains five items referring to “selfreliance”
(1, 5, 7, 12, and 14), three items referring to “meaningfulness” (2, 9, and 13), two items
referring to “equanimity” (3 and 10), two items referring to “perseverance” (6 and 8), and
two items referring to “existential aloneness” (4 and 11). The participants rate the items
on a scale from 1 (strongly disagree) to 7 (strongly agree).
12Item PurposeinLife Test (PILTest12). The PILTest12 (Aquino et al., 2009) is
a reduced version of the Purpose in Life Test, developed by Crumbaugh and Maholick
(1964). Composed of 12 items, the PILTest12 evaluates, through a 7point Likert scale
(1 5 totally disagree and 7 5 totally agree), the degree of one’s meaning in life. The
higher the score, the higher the sense of meaning in life. In previous studies (Aquino,
2009; Aquino et al., 2009), the PILTest12 presented a reliable singlefactor solution,
attested by Cronbach’s a 5 .83. In this study, Cronbach’s alpha was .82.
General Health Questionnaire (GHQ12). The GHQ12 is the reduced version of
the original General Health Questionnaire (Goldberg, 1972) and is one of the most
widely used instruments to evaluate the degree of one’s psychological wellbeing. The
respondents rate the questionnaire on a scale ranging from 1 (more than habitual) to
4 (less than habitual). Brazilian validation studies reported a twofactor solution as the
most reliable, with Cronbach’s alpha ranging from .85 to .63 (Gouveia, Barbosa, Andrade,
& Carneiro, 2010; Sarriera, Schwarcz, & Câmara, 1996). In this study, Cronbach’s alpha
was .80 and .66, to “depression” and “selfefficacy” subscales, respectively.
Data Analysis
The total sample was randomly split in two halves to analyze the construct validity of
the RS14. Different exploratory factor analyses (EFAs) were conducted with the first
half (n 5 567), using PCA, according to the previous studies, and maximum likelihood
(ML) extraction method. The first EFA was performed because of the importance on
conducting the analysis with the same methods used in the previous and original studies
(Nishi et al., 2010; Wagnild, 2010). However, considering that PCA is only a data
reduction method, based on the linear correlation among the measured variables, and
it is not appropriate when the objective is to identify latent constructs underlying a set
of measured variables (Costello & Osborne, 2005), a second estimation method, ML,
was chosen. For all EFAs, the oblimin rotation was chosen because of the possibility to
allow factor correlations (Fabrigar, Wegener, MacCallum, & Strahan, 1999). The sample
adequacy was assessed by the KaiserMeyerOlkin (KMO) and Bartlett’s sphericity test
measures. Reliability was assessed using Cronbach’s alpha index. The number of factors
extracted in the EFAs was confronted with Monte Carlo parallel analyses, which was
used as the principal guideline to the factor retention, because of its better accuracy
(Hayton, Allen, & Scarpello, 2004).
Copyright © Springer Publishing Company, LLC
Page 6
Psychometric Properties of Brazilian RS14 135
Different CFAs with the second half of the sample (n 5 572), using ML estimation
method, were conducted in order to evaluate the goodness of fit of the exploratory models.
The ML estimation method was chosen, for both EFA and CFA, once our data presented
an acceptable (i.e., nonsevere) degree of normality deviation (Lei & Lomax, 2005; Olsson,
Foss, Troye, & Howell, 2000; West, Finch, & Curran, 1995).
The absolute fit indexes calculated were chisquare (x2), chisquare/degree of freedom
ratio (x2/df), and standardized root mean square residual (SRMR). Parsimony fit was the
root mean square error of approximation (RMSEA). The comparative indexes used were
comparative fit index (CFI) and TuckerLewis index (TLI). These indexes were chosen
because of their favorable performance in Monte Carlo researches (Brown, 2006; Hu &
Bentler, 1999). According to many guidelines, x2/df might be less than 2 or 3; the closer
SRMR to 0 the better; RMSEA values less than .06 indicate a good fit, between .06 and
.08 a reasonable fit, between .08 and .10 a mediocre fit, and more than .10 a poor fit. CFI
and TLI must be higher than .90 or .95 or close to it (Brown, 2006; Byrne, 2010). Both
samples were appropriate to use the factor analyses according to multiple criteria: They
include more than 200 subjects; exceed the sample/item ratio of 10:1 for the EFA (Hair et
al., 2006); and exceed the ratio of 10 subjects for each parameter to be estimated, for the
CFA (Brown, 2006; Byrne, 2010).
Multigroup CFA (MGCFA) analyses were then conducted to evaluate measurement
invariance of the RS14 through distinct age (youngsters: from 14 to 29 years old,
n 5 762; and adults: from 30 to 59 years old, n 5 377) and gender (male, n 5 423 and
female, n 5 716) groups. The age groups aforementioned were defined according to the
Brazilian Institute of Geography and Statistic (IBGE) guidelines (IBGE, 1999). Although
the range from 14 to 29 years old encompasses three different levels of youth (adolescent,
young, and matureyoung), these subdivisions were not accessed in this study.
For both MGCFA (age and gender), four different models are presented. Model 1
(equal number of factors or configural invariance) is an unconstrained model and assesses
whether the number of factors and the pattern of fixed and free parameters are equal across
groups. Model 2 (equal factor loadings or metric invariance) analyzes if the factor loadings
are equal across groups (i.e., it determines whether the measures have the same meaning
and structure for different groups of respondents). Model 3 (equal latent variance or struc
tural invariance) investigates whether latent (co)variance are equal across groups. Finally,
Model 4 (equal measurement residuals or strict factorial invariance) evaluates whether
the measurement residuals are equal across groups. The levels of assessment are ordered
hierarchically, from Model 1 to Model 4. Thus, each constrained model is nested within
a less restricted one (Cheung & Rensvold, 2002). Differences between the models were
evaluated by the chisquare difference test (Dx2) and CFI difference test (DCFI; Cheung
& Rensvold, 2002; Vandenberg & Lance, 2000). Significant differences (i.e., Dx2[df] sig
nificant at p . .05 and DCFI . .01) observed between the goodness of fit indexes of the
models indicate that the factor parameters are not the same across the specified groups.
Convergent validity of the RS14 was performed using the PILTest12 and the GHQ12
in a random sample of 250 subjects. We expected that the RS14 correlated positively
with the PILTest12 (meaning in life) and GHQ12 (selfefficacy) and negatively with the
GHQ12 (depression). MannWhitney tests with Monte Carlo simulations (99% confi
dence interval; 10,000 random samples) were conducted to analyze the effects of age and
gender on the RS scores. The nonparametric tests were chosen instead of Student’s t test
because of the large nonequivalence on the group sizes, which could affect the obtained
results (Field, 2005; Markowski & Markowski, 1990).
Copyright © Springer Publishing Company, LLC
Page 7
136 Damásio et al.
RESULTS
Exploratory Factor Analyses
The first EFA, using PCA, presented a onecomponent solution, which accounted for
31.93% of the explained variance of the construct (KMO 5 .874; Bartlett’s test of
sphericity x2[91] 5 1,755.466, p , .001). All the 14 items loaded on the first component,
with satisfactory factor loadings (i.e., . .30). For this solution, Cronbach’s alpha was .82.
A second EFA using ML extraction method with oblimin rotation encountered a two
factor solution with eigenvalue . 1 (Factor I: Eigenvalue 5 4.46; Factor II: Eigenvalue 5
1.13). The first factor was composed of 8 items (2, 3, 4, 5, 10, 11, 12, and 13), whereas the
second factor was composed of 6 items (1, 6, 7, 8, 9, and 14). For both factors, Cronbach’s
alpha was .73. The adequacy of this twofactor solution was assessed through a parallel
analysis using marginally bootstrapped samples (PA–MBS; Lattin, Carroll, & Green,
2003). The results showed that a singlefactor solution is the most appropriate (Figure 1).
When a single factor solution was forced, using the ML estimation method, item 3
(. . . Take things in stride) did not load significantly (i.e., . .30). For this model composed of
13 items, Cronbach’s alpha was .83, and the explained variance was 31.93% (Table 1).
Confirmatory Factor Analyses and Multigroup Comparisons
Different CFAs with the second half of the sample (n 5 572) were conducted in order
to assess the fit indexes of the obtained exploratory models. The goodness of fit using
absolute, parsimony, and comparative fit indexes, as recommended by Brown (2006), are
presented in Table 2.
Factor Number
6
5
4
3
2
1
0
12345678910 111213 14
Eigenvalues
Dataset Eigenvalues
Random Eigenvalues
Figure 1. Parallel analysis using marginally bootstrapped samples.
Copyright © Springer Publishing Company, LLC
Page 8
Psychometric Properties of Brazilian RS14 137
TABLE 1. Exploratory Factor Analyses (PCA and ML) and Reliability Coefficients
of the Brazilian RS14 (n 5 567)
Items (Short Content)
Extraction Methods
Principal
Component Analysis
Resilience
Maximum
Likelihood
Resilience
9 . . . Be interested in things
2 . . . Have accomplished things in life
6 . . . To be determined
14 . . . Find my way out of difficulties
4 . . . Friends with myself
11 . . . Belief on itself
13 . . . Meaning in life
12 . . . People can rely on me
8 . . . Be selfdisciplined
5 . . . Handle many things
1 . . . Manage things one way or another
7 . . . Get through difficult times
10 . . . Find something to laugh about
3 . . . Take things in stride
.68a
.66a
.64a
.63a
.63a
.60a
.58a
.57a
.56a
.52a
.50a
.49a
.45a
.31a
.64a
.62a
.61a
.58a
.58a
.55a
.53a
.51a
.52a
.46a
.44a
.43a
.40a
.27
Eigenvalue
% explained variance
M
SD
Cronbach’s alpha coefficient
4.47
31.93
76.27
12.71
.82
4.47
31.93
72.24
10.20
.83
aItems loaded significantly in the factor.
TABLE 2. Goodness of fit for the Resilience Scale
Models
Goodness of Fit Indexes
x2 (df)
x2/df
SRMRRMSEA (90% CI) CFITLI
RS14 (PCA)
RS13 (ML)
213.820 (77)
192.180 (65)
2.777
2.957
.042
.041
.056 (.047–.065)
.059 (.049–.065)
.924
.928
.910
.913
Note. SRMR 5 standardized root mean square residual; RMSEA 5 root mean square
error of approximation; CI 5 confidence interval; CFI 5 comparative fit index;
TLI 5 TuckerLewis index; PCA 5 principal component analysis; ML 5 maximum
likelihood.
Copyright © Springer Publishing Company, LLC
Page 9
138 Damásio et al.
The goodness of fit for the different models were quite similar. As it can be seen in
Table 2, the ratio x2/df and the RMSEA indexes of the PCA model are slightly better than
the ML onefactor model. However, the SRMR, CFI, and TLI indexes of the ML one
factor model are better than the PCA one. The overall good fit of the models suggested
the possibility of conducting measurement invariance analyses (Brown, 2006). For the
measurement invariance and subsequent analyses, the RS13 was chosen, considering the
better adequacy of the extraction method.
MGCFA were then conducted in the RS13 to assess whether the parameters of
this obtained factorial model were invariant across gender (male and female) and age
( youngsters and adults) groups (Table 3). The first MGCFA was conducted in the gender
group. The goodness of fit of Model 1 (equal number of factors) demonstrated, by the
acceptable goodness of fit indexes, that the posited exploratory model is plausible for
both men and women (i.e., same items measuring the same latent construct), matching
the criteria of configural invariance (Byrne, 2010; Cheung & Rensvold, 2002). Model 2
(equal factor loadings), which evaluated the assumption of metric invariance, also had an
overall good fit to the data and did not significantly degrade fit relative to the first model,
Dx2(12) 5 19.750, ns (critical value for x2[12] 5 21.03, a 5 .05), and DCFI , .01. More
than that, the second model demonstrated a slight improvement in the RMSEA and TLI
goodness of fit indexes, comparing to the first one. Because the constraint of equal factor
loadings did not significantly degrade the fit of the RS13, it can be concluded that the
indicators (items) evidence comparable relationships to the latent constructs for both male
and female (Brown, 2006; Byrne, 2010).
Considering the same criteria aforementioned (Dx2, DCFI), the fit of Model 3 (equal
latent variance) presented contradictory findings. Although the Dx2 indicates significant
changes in the model, Dx2(1) 5 19.830 (critical value for Dx2[1] 5 3.84, a 5 .05), the
DCFI indicate that the models did not change significantly (DCFI 5 .006). Model 4 (equal
measurement residuals) presented a significant degradation of the model as evaluated by
the chisquare distribution, Dx2(13) 5 57.009, p , .05 (critical value for Dx2[13] 5 22.36,
a 5 .05) and by the CFI difference test (DCFI 5 .013). In other words, the assumption of
equal measurement residuals could not be achieved.
A second MGCFA was conducted in order to evaluate the RS13 measurement
invariance across two different age groups: youngsters and adults (Table 3). Model 1
(equal number of factors) presented acceptable goodness of fit indexes, suggesting that
the factor structure of the RS13 is equal for both youngsters and adults groups. Model
2 (equal factor loadings) showed an overall good fit to the data and did not significantly
degrade fit relative to the first model, Dx2(12) 5 12.014, ns (critical value for x2[12] 5
21.03, a 5 .05), and DCFI , .01. Relative to Model 3 (equal latent variance), once again,
contradictory results were found: The Dx2 indicated significant changes in the model,
Dx2(1) 5 8.605 (critical value for Dx2[1] 5 3.84, a 5 .05), but the DCFI indicated
that the model did not change significantly (DCFI 5 .002). Finally, regarding Model 4
(equal measurement residuals), the assumption of invariance over again could not be
achieved, neither by the chisquare difference tests, Dx2(13) 5 376,607 (critical value for
Dx2[13] 5 22.36, a 5 .05) nor by the CFI difference test (DCFI . .01).
Convergent Validity
Pearson’s correlations among the RS13 with the meaning in life (PILTest12) and depres
sion and selfefficacy (GHQ12), in a random sample of 250 subjects, were performed.
Copyright © Springer Publishing Company, LLC
Page 10
Psychometric Properties of Brazilian RS14 139
TABLE 3. Fit Indexes for Gender and Age MGCFA for the RS13RS13
Goodness of Fit Indexes
x2
df
x2/df
Dx2
Ddf
RMSEA (90% CI)
SRMR
CFI
TLI
Gender measurement invariance
Equal number of factors
416.609
130
3.189


.044 (.039–.049)
.041
.915
.894
Equal factor loadings
434.359
142
3.059
19.750
12
.043 (.038–.047)
.046
.913
.904
Equal latent variance
454.189
143
3.176
19.830
1
.044 (.039–.048)
.069
.907
.899
Equal measurement residuals
511.198
156
3.277
57.009
13
.045 (.040–.049)
.066
.894
.894
RS13
Age measurement invariance
Equal number of factors
381.449
130
2.934


.041 (.036–.046)
.046
.925
.910
Equal factor loadings
393.463
142
2.771
12.014
12
.039 (.035–.044)
.048
.925
.917
Equal latent variance
402.068
143
2.812
8.605
1
.040 (.035–.045)
.063
.923
.915
Equal measurement residuals
778.675
156
4.992
376.607
13
.059 (.055–.063)
.058
.814
.814
Note. RS 5 resilience scale; MGCFA 5 multigroup confirmatory factor analysis; SRMR 5 standardized root mean square
residual; RMSEA 5 root mean square error of approximation; CI 5 confidence interval; CFI 5 comparative fit index;
TLI 5 TuckerLewis index.
Copyright © Springer Publishing Company, LLC
Page 11
140 Damásio et al.
As expected, the RS13 correlated positively with meaning in life and selfefficacy and
negatively with depression (see Table 4).
Effects of Age and Gender on the RS13
Performance differences on the RS13 were compared between two different groups
(gender, men and women; and age, youngsters and adults) using MannWhitney test
with Monte Carlo simulations. No differences between males (Mdn 5 73) and females
(Mdn 5 74) were found (U 5 143618.50, z 5 21.458, p . .10). However, the youngsters
group presented a lower level of resilience (Mdn 5 72), when compared to the adults
group (Mdn 5 76, U 5 121524.00, z 5 24.236, p , .001).
DISCUSSION
Initially, we conducted two different EFAs, with the first half of the sample, in order to
evaluate the possibility of different factor solutions for the RS14. The first EFA, using
PCA, was conducted according to the guidelines presented in a previous (Nishi et al.,
2010) and original validation study (Wagnild, 2010). The results presented a onefactor
solution, with a Cronbach’s a 5 .82. All 14 items loaded higher than .30. When a second
extraction method (ML with oblimin rotation) was chosen, the RS14 presented initially
a clear twofactor solution (i.e., without crossloadings). However, parallel analysis
presented a singlefactor solution as the most reliable. When a singlefactor solution was
forced, using the ML extraction method, the Item 3 (. . . Take things in stride) did not load
significantly. It is important to consider that PCAs does not differentiate the common and
unique variance of the items, which tends to increase the component loadings (Fabrigar
et al., 1999). Probably because of this, Item 3 loaded significantly in the PCA solution
but not on the ML one. Considering that the previous reported studies (Wagnild, 2010;
Nishi et al., 2010) used only PCA, we do not know if the nonsignificant loading of Item
3, using a factor analytic method is a peculiarity of our sample or not. Although Item 3
TABLE 4. Means, Medians, Standard Deviations, Reliabilities, and Correlation
Matrix of Measured Variables
1234
1. Resilience
2. Meaning in life
3. Selfefficacy
4. Depression
(.83)
.55**
.46**
2.47**
(.82)
.41**
2.37**
(.66)
2.46**(.80)
M (possible range)
Mdn
SD
Note. 1 5 Resilience (RS13); 2 5 Meaning in life (PILTest12); 3 5 Selfefficacy
(GHQ12); 4 5 Depression (GHQ12). Values in parenthesis along the diagonal
represent internal consistency estimates (Cronbach’s alpha).
** p , .001.
74.5 (7–91)
76.0
8.94
68.9 (7–84)
71
10.47
15.7(5–20)
16.0
4.80
16.2 (7–28)
19.0
2.30
Copyright © Springer Publishing Company, LLC
Page 12
Psychometric Properties of Brazilian RS14 141
loaded well in Wagnild’s study (factor loading 5 .46), further studies (Brazilians and
internationals) considering different extraction methods are encouraged to better clarify
this initial finding.
For the 13item model, Cronbach’s a 5 .83, whereas the 14item model presented
a Cronbach’s alpha of .82. Both PCA and ML singlefactor solutions presented a poor
amount of explained variance (Peterson, 2000), similar to the results found by Nishi et al.
(2010).
In order to identify the better fit to the Brazilian RS14, two different CFAs were
conducted, testing the exploratory models. The results presented good indexes for both
models, suggesting similar plausibility. However, MGCFA were conducted in the factor
analytic model (i.e., the ML 13item solution) in order to evaluate measurement invariance
on age and gender groups.
For the gender and age groups, the assumptions of “equal number of factors”
(configural invariance) and “equal factor loadings” (metric invariance) were totally
accepted, suggesting that the structure (number of factors and disposition of the items) and
the factor loadings were similar for male and female, as well as for youngsters and adults.
The assumption of “equal latent variance” (structural invariance) presented contradictory
findings in both groups. The chisquare difference test (Dx2) indicated significant changes
in the model, whereas the CFI difference test (DCFI) attested the contrary. It is known,
however, that chisquare difference tests between nested models are sample size dependent
(Brown, 2006; Meade, Johnson, & Braddy, 2006) and sensitive for models with several
constrained parameters (Marsh, Balla, & McDonald, 1988; Vandenberg & Lance, 2000).
In turn, the DCFI metric is considered a robust fit statistic when testing MGCFA models
(Cheung & Rensvold, 2002; Meade et al., 2006). So, considering the DCFI results, it is
possible to argue that the assumption of equal latent variance was achieved for both groups.
Finally, the assumption of “equal measurement residuals” (strict factorial invariance)
could not be achieved neither for the gender nor for the age group. However, it is widely
accepted that testing measurement residuals invariance represents an overly restrictive test
of the data on psychological research and does not compromise the conclusions about the
measurement invariance (Byrne, 2010).
Regarding the convergent validity, the Brazilian RS13 correlated positively with meaning
in life (PILTest12) and selfefficacy (GHQ12) and negatively with depression (GHQ12).
These results are consistent with previous studies. Nishi et al. (2010), for example, found
that the RS14 was negatively correlated with depression (r 5 2.28) and positively cor
related with selfesteem (r 5 .28). As largely discussed, meaning in life, selfefficacy, and
selfesteem are positive dispositional aspects that are associated with higher likelihood of
resilient responses to a variety of life stressors (Masten & Reed, 2002; Moskowitz, 2010).
Depression, in turn, has been considered a vulnerability factor that is generally negatively
correlated with resilience (Zautra, Hall, & Murray, 2010). Thus, the Brazilian RS13 pre
sented a significant convergent validity.
Consistent with previous studies using the RS, the levels of resilience increased with
age and presented no relation with gender (Lundman, Strandberg, Eisemann, Gustafson,
& Brulin, 2007; Nishi et al., 2010; Portzky, Wagnild, De Bacquer, & Audenaert, 2010;
Wagnild, 2010). Lundman et al. (2007) argued that resilience is not static and probably
is a process developed during the life span. According to these authors, despite the fact
that old age has been described as a period of physical, functional, and social losses,
other age groups as youngsters and adults might have an increasing resilience levels with
age, probably influenced by individual and contextual factors. One important aspect to
Copyright © Springer Publishing Company, LLC
Page 13
142 Damásio et al.
consider, however, is that none of these reported studies, including this one, evaluated
the levels of resilience in atrisk populations but only in healthy or general citizens.
Moreover, resilience as measured by the RS14 is conceptualized as a personality trait
(Wagnild & Young, 1993), and maybe this specificity of the instrument is influencing the
observed results regarding age and resilience levels. So, these results must be cautiously
interpreted.
Two main limitations of this study are pointed out. First, the sample was not paired
by age, and this can be seen as a limitation for group comparisons. Although the groups
of youngsters and adults were based on a very clear Brazilian criterion, the range from
14 to 29 (youngsters) and 30 to 59 (adults) is very large and probably specificities on the
degree of resilience according to developmental characteristics could not be well accessed.
Other important aspect to highlight is the fact that our sample was composed by a general
population. To include and to identify groups in atrisk situations is a very important
issue for further studies, especially considering the construct that the scale is proposed to
measure (resilience).
CONCLUSIONS
The objective of this study was to present the factor structure and the psychometric
properties of the RS14 in the Brazilian context. EFAs and CFAs presented a reliable
and plausible onefactor solution. The onefactor solution using PCA presented a similar
structure as reported in previous studies. However, using a true factor analysis extraction
method (ML), Item 3 did not load significantly.
MGCFA with the ML onefactor solution presented satisfactory results, testifying that
the factor structure (number of factors and items) and some measurement parameters
(i.e., factor loadings and latent variance) were invariant across gender and age groups.
These results suggest that the Brazilian RS13 seems to be a reliable measure to be used
in the Brazilian general population.
Convergent validity was also satisfactory evaluated. The RS13 correlated negatively
with depression and positively with meaning in life and selfefficacy. MannWhitney’s
test replicated previous findings, suggesting that resilience, as measured by the RS13,
increases with age. However, studies with different subjects (e.g., atrisk populations,
different socioeconomic status) are necessary to better understand these results. The fact
that the level of resilience increased with age is not a contradictory finding related to the
measurement invariance of the instrument, once the scores and means of the items were
not evaluated by the MGCFA. Finally, it is suggested that further investigations should be
conducted in other regions of the country, considering also sociodemographic, economics,
familiar, and other contextual aspects, in order to replicate or not these findings.
REFERENCES
Ahern, N. R., Kiehl, E. M., Sole, M. L., & Byers, J. (2006). A review of instruments measuring
resilience. Issues in Comprehensive Pediatric Nursing, 29(2), 103–125.
Aquino, T. A. A. (2009). Atitudes e intenções de cometer suicídio: seus correlatos existenciais e
normativos. Unpublished doctoral dissertation, Federal University of Paraíba, João Pessoa,
Brazil.
Copyright © Springer Publishing Company, LLC
Page 14
Psychometric Properties of Brazilian RS14 143
Aquino, T. A. A., Correia, A. P. M., Marques, A. L. C., Souza, C. G., Freitas, H. C. A., Araújo, I. F.,
et al. (2009). Atitude religiosa e sentido de vida: Um estudo correlacional [Religious attitude and
the meaning of life: A correlational study]. Psicologia: Ciência e Profissão, 29(2), 228–243.
Brown, T. A. (2006). Confirmatory factor analysis for applied research. New York: The Guilford
Press.
Byrne, B. M. (2010). Structural equation modeling with AMOS: Basic concepts, applications, and
programming (2nd ed.). New York: Routledge, Taylor & Francis.
CampbellSills, L., & Stein, M. B. (2007). Psychometric analysis and refinement of the
Connor–Davidson Resilience Scale (CDRISC): Validation of a 10item measure of resilience.
Journal of Traumatic Stress, 20(6), 1019–1028.
Charney, D. S. (2004). Psychobiological mechanisms of resilience and vulnerability: Implications
for successful adaptation to extreme stress. The American Journal of Psychiatry, 161(2),
195–216.
Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodnessoffit indexes for testing measure
ment invariance. Structural Equation Modeling, 9(2), 233–255.
Connor, K. M., & Davidson, J. R. (2003). Development of a new resilience scale: The Connor–Davidson
Resilience Scale (CDRISC). Depression and Anxiety, 18(2), 76–82.
Costello, A. B., & Osborne, J. W. (2005). Best practices in exploratory factor analysis: Four
recommendations for getting the most from your analysis. Practical Assessment, Research &
Evaluation, 10(7), 1–9.
Crumbaugh, J. C., & Maholick, L. T. (1964). An experimental study in existentialism: The psycho
metric approach to Frankl’s concept of noogênic neurosis. Journal of Clinical Psychology,
20(1), 200–207.
Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating the use of
exploratory factor analysis in psychological research. Psychological Methods, 4(3), 272–299.
Field, A. P. (2005). Discovering statistics using SPSS (2nd ed.). London: Sage.
Friborg, O., Barlaug, D., Martinussen, M., Rosenvinge, J. H., & Hjemdal, O. (2005). Resilience
in relation to personality and intelligence. International Journal of Methods in Psychiatric
Research, 14(1), 29–42.
Goldberg, D. P. (1972). The detection of psychiatric illness by questionnaire: A technique for the
identification and assessment of nonpsychotic psychiatric illness. London: Oxford University
Press.
Gouveia, V. V., Barbosa, G. A., Andrade, E. O., & Carneiro, M. B. (2010). Factorial validity and
reliability of the General Health Questionnaire (GHQ12) in the Brazilian physician population.
Cadernos de Saúde Pública, 26(7), 1439–1445.
Hair, J. F., Black, B., Babin, B., Anderson, R. E., Tatham, R. L., & Black, W. C. (2006). Multivariate
data analysis (6th ed.). Englewood Cliffs, NJ: Prentice Hall.
Hayton, J. C., Allen, D. G., & Scarpello, V. (2004). Factor retention decisions in exploratory factor
analysis: A tutorial on parallel analysis. Organizational Research Methods, 7(2), 191–205.
Heilemann, M. V., Lee, K., & Kury, F. S. (2003). Psychometric properties of the Spanish version of
the Resilience Scale. Journal of Nursing Measurement, 11(1), 61–72.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis:
Conventional criteria versus new alternatives. Structural Equation Modeling, 6(1), 1–55.
Instituto Brasileiro de Geografia e Estatística. (1999). Síntese de Indicadores Sociais [Synthesis of
social indicators]. Rio de Janeiro, Brazil: Instituto Brasileiro de Geografia e Estatística.
Lattin, J. M., Carroll, D. J., & Green, P. E. (2003). Analyzing multivariate data (pp. 114–116).
Belmont, CA: Duxbury Press.
Lei, M., & Lomax, R. G. (2005). The effect of varying degrees of nonnormality in structural equation
modeling. Structural Equation Modeling, 12(1), 1–27.
Lundman, B., Strandberg, G., Eisemann, M., Gustafson, Y., & Brulin, C. (2007). Psychometric
properties of the Swedish version of the Resilience Scale. Scandinavian Journal of Caring
Sciences, 21(2), 229–237.
Luthar, S. S., Cicchetti, D., & Becker, B. (2000). The construct of resilience: A critical evaluation
and guidelines for future work. Child Development, 71(3), 543–562.
Markowski, C. A., & Markowski, E. P. (1990). Conditions for the effectiveness of a preliminary test
of variance. The American Statistician, 44(4), 322–326.
Copyright © Springer Publishing Company, LLC
Page 15
144 Damásio et al.
Marsh, H. W., Balla, J. R., & McDonald, R. P. (1988). Goodnessoffit indexes in confirmatory
factor analysis: The effect of sample size. Psychological Bulletin, 103(3), 391–410.
Masten, A. S. (2001). Ordinary magic: Resilience processes in development. The American
Psychologist, 56(3), 227–238.
Masten, A. S., Best, K. M., & Garmezy, N. (1990). Resilience and development: Contributions
from the study of children who overcome adversity. Development and Psychopathology, 2(4),
425–444.
Masten, A. S., & Reed, M. J. (2002). Resilience in development. In C. R. Snyder & S. J. Lopez
(Eds.), Handbook of positive psychology (pp. 74–88). Oxford, United Kingdom: Oxford
University Press.
Masten, A. S., & Wright, M. O. (2010). Resilience over the lifespan: Developmental perspectives
on resistance, recovery, and transformation. In J. W. Reich, A. J. Zautra, & J. S. Hall (Eds.),
Handbook of adult resilience (pp. 213–237). New York: The Guilford Press.
Meade, A. W., Johnson, E. C., & Braddy, P. W. (2006, August). The utility of alternative fit indices
in tests of measurement invariance. Paper presented at the annual Academy of Management
conference, Atlanta, GA.
Moskowitz, J. T. (2010). Positive affect at the onset of chronic illness: Planting the seeds of
resilience. In J. W. Reich, A. J. Zautra, & J. S. Hall (Eds.), Handbook of Adult Resilience
(pp. 465–483). New York: The Guilford Press.
Nishi, D., Uehara, R., Kondo, M., & Matsuoka, Y. (2010). Reliability and validity of the Japanese
version of the Resilience Scale and its short version. BMC Research Notes, 3(1), 310.
Nygren, B., Randström, K. B., Lejonklou, A. K., & Lundman, B. (2004). Reliability and validity of
a Swedish language version of the Resilience Scale. Journal of Nursing Measurement, 12(3),
169–178.
Olsson, U. H., Foss, T., Troye, S. V., & Howell, R. D. (2000). The performance of ML, GLS, and
WLS estimation in structural equation modeling under conditions of misspecification and non
normality. Structural Equation Modeling, 7(4), 557–595.
Oshio, A., Kaneko, H., Nagamine, S., & Nakaya, M. (2003). Construct validity of the Adolescent
Resilience Scale. Psychological Reports, 93(3 Pt. 2), 1217–1222.
Pesce, R. P., Assis, S. G., Avanci, J. Q., Santos, C. N., Malaquias, J. V., & Carvalhaes, R. (2005).
Adaptação transcultural, confiabilidade e validade da escala de resiliência [Crosscultural
adaptation, reliability and validity of the resilience scale]. Cadernos de Saúde Pública, 21(2),
436–448.
Peterson, R. A. (2000). A metaanalysis of variance accounted for and factor loadings in exploratory
factor analysis. Marketing Letters, 11(3), 261–275.
Portzky, M., Wagnild, G., De Bacquer, D., & Audenaert, K. (2010). Psychometric evaluation of the
Dutch Resilience Scale RSnl on 3265 healthy participants: A confirmation of the association
between age and resilience found with the Swedish version. Scandinavian Journal of Caring
Sciences, 24(Suppl. 1), 86–92.
Rodríguez, M., Pereyra, M. G., Gil, E., Jofré, M., De Bortoli, M., & Labiano, L. M. (2009).
Propiedades psicométricas de la escala de resiliencia versión argentina. Evaluar, 9(1), 72–82.
Ryan, L., & Caltabiano, M. L. (2009). Development of a new resilience scale: The Resilience in
Midlife Scale (RIM Scale). Asian Social Science, 5(11), 39–51.
Sarriera, J. C., Schwarcz, C., & Câmara, S. G. (1996). Bemestar psicológico: Análise fatorial da
escala de Goldberg (GHQ12) numa amostra de jovens. Psicologia: Reflexão e Crítica, 9(2),
293–306.
Sinclair, V. G., & Wallston, K. A. (2004). The development and psychometric evaluation of the Brief
Resilient Coping Scale. Assessment, 11(1), 94–101.
Smith, B. W., Dalen, J., Wiggins, K., Tooley, E., Christopher, P., & Bernard, J. (2008). The brief
resilience scale: Assessing the ability to bounce back. International Journal of Behavioral
Medicine, 15(3), 194–200.
Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement invari
ance literature: Suggestions, practices, and recommendations for organizational research.
Organizational Research Methods, 3(1), 4–70.
von Soest, T., Mossige, S., Stefansen, K., & Hjemdal, O. (2010). A validation study of the Resilience
Scale for Adolescents (READ). Journal of Psychopathology and Behavioral Assessment, 32(2),
215–225.
Copyright © Springer Publishing Company, LLC