The Personal Wellbeing Index: Psychometric ...

Soc Indic Res (2013) 110:913–924 DOI 10.1007/s11205-011-9964-9

The Personal Wellbeing Index: Psychometric Equivalence for Adults and School Children Adrian J. Tomyn • Matthew D. Fuller Tyszkiewicz Robert A. Cummins

•

Accepted: 14 November 2011 / Published online: 27 November 2011 Ó Springer Science+Business Media B.V. 2011

Abstract Despite the wealth of accumulated research evaluating subjective wellbeing (SWB) in children and adults, the validity of scores from parallel forms of SWB measures for each age group has yet to be empirically tested. This study examines the psychometric equivalence of the child and adult forms of the personal wellbeing index (PWI) using multiple-group confirmatory factor analysis. The child sample comprised 1,029 Victorian high-school students (aged 11–20) sampled across three independent studies. The adult sample comprised 1,965 Australian adults drawn from the Australian Unity Wellbeing Index. The results demonstrated strict factorial invariance between both versions, suggesting that the PWI measures the same underlying construct in adolescent and adult populations. These findings provide support for quantitative comparisons between adult and adolescent SWB data as valid. Keywords Personal wellbeing index Psychometric equivalence Adolescents Adults Confirmatory factor analysis

1 Introduction A considerable body of literature has accumulated concerning the subjective wellbeing (SWB) of both adults and adolescents. However, the scales used for each group are generally age-specific, so the extent to which they measure the same construct is uncertain. This is a crucial gap in understanding how SWB changes across the lifespan. A. J. Tomyn (&) Discipline of Psychology, School of Health Sciences, RMIT University, Bundoora, Australia e-mail: [email protected] M. D. Fuller Tyszkiewicz R. A. Cummins Deakin University, Melbourne, Australia e-mail: [email protected] R. A. Cummins e-mail: [email protected]

123

914

A. J. Tomyn et al.

The International Wellbeing Group (2011) was formed to address the issue of valid and reliable cross-group measurement of SWB. The instrument under investigation by the Group is the personal wellbeing index (PWI; IWG 2006), which seems to have promise in this regard. The PWI comprises eight domains (standard of living, health, achieving in life, relationships, safety, community-connectedness, future security, and religion/spirituality) which represent the first level deconstruction of the global question, ‘How satisfied are you with your life as a whole?’ It has been established that these domains are strongly related to global life satisfaction (GLS) in western and non-western cultures, including Australia, China, and Algeria (IWG 2006). One theory of SWB that describes consistency and change across the lifespan is Homeostasis Theory (Cummins 1995, 2010; Cummins and Nistico 2002; Davern et al. 2007). This theory is based on the premise that each person has a biologically determined level of SWB maintained within a positive, narrow range of values around a set-point. The hypothesised average SWB set-point of approximately 75% points (Cummins 1995, 2010; Cummins and Nistico 2002) is supported by data obtained as part of the Australian Unity Wellbeing Index. Over the period 2001–2010, 24 surveys of the Australian adult population were conducted (Cummins et al. 2010). Using the survey mean SWB scores as data, the normative range for SWB is between 73.7 and 76.7% points. This remarkable stability, which has also been observed at the level of individuals (e.g. Headey and Wearing 1989, 1992), is accounted for by homeostasis theory. According to this theory, homeostasis seeks to defend individual SWB set-points, which are argued to reflect the affective core of SWB, termed homeostatically protected mood (HPMood; Cummins 2010). It is proposed that this genetically determined, affective mood state is protected by both external and internal buffers (see Cummins 2010 for a review). Importantly, this theory asserts that the homeostatic defence of SWB is common to all people and across the human lifespan. However, tests of this claim require measures that function equivalently in adult and child populations. 1.1 The Personal Wellbeing Index-School Children (PWI-SC) Scales with acceptable psychometric properties for adults usually need to be modified for children or people with an intellectual disability, in order to accommodate lower levels of reading and comprehension (Cummins et al. 2004). In this vein, the personal wellbeing index-school children (PWI-SC; Cummins and Lau 2005) was created to allow for valid measurement of SWB for adolescents. Items comprising the PWI-SC are based on the adult version of this scale and have been simplified for this group. For example, the item ‘How satisfied are you with your future security?’ is simplified to ‘How satisfied are you about what may happen to you later on in your life?’ The PWI-SC also asks respondents to indicate their level of ‘happiness’ rather than ‘satisfaction’ as in the adult version, but using the same 0–10 end-defined scale as for adults. The adjective ‘happy’ is argued to be less abstract and more comprehensible for school age adolescents (Cummins and Lau 2005). However, the practice of scale modification raises concerns about the relationship of scores to norms based on the adult scale (Cummins et al. 2004). It also raises questions about the comparable validity of parallel versions. In an attempt to address these issues, Tomyn and Cummins (2011a) evaluated the psychometric properties of a modified version of the PWI-SC (PWI-SCb) that incorporated the adjective ‘satisfied’ to increase uniformity with the adult scale in a sample of 351 Australian high school students. Principal Components Analysis revealed that the scale was adequately fit by a one-factor solution, and that factor loadings for the domains were strong, ranging from 0.53 (Community) to 0.76 (Health). Furthermore, and consistent with Australian adult normative data and the

123

The Personal Wellbeing Index

915

prediction of SWB Homeostasis theory, the student mean SWB of 74.7 points was within the Australian adult normative range (73.7–76.7% points; Cummins et al. 2010). Additionally, in terms of comparable predictive validity, Tomyn and Cummins (2011b) determined that SWB and the related construct of HPMood are similarly related in adults and children. There were also some differences in the functioning of the PWI-SCb and PWI-A. First, although most of the domains of the PWI-SCb had means within the adult normative range, Tomyn and Cummins (2011a) reported several exceptions. The mean for the domain of ‘Achieving in life’ was significantly lower in the adolescent sample (p \ 0.001), while the mean for ‘Community connectedness’ was significantly higher (p \ 0.05). Second, some differences were observed in the profile of domains that predict GLS. For adults, all domains except ‘Safety’ contribute unique variance to prediction of GLS in Australia (IWG 2006). For adolescents, on the other hand, ‘Safety’ contributed unique variance while the domains of ‘Relationships’ and ‘Community connectedness’ did not. These different content validity profiles may reflect substantive differences in the expression and function of SWB from adolescence to adulthood. It is also possible that simplifications made to both versions of the PWI-SC, to accommodate an adolescent audience, contribute measurement error. One method to investigate this issue is the statistical technique of measurement invariance. This can be used to disambiguate the joint effects of measurement biases and substantive differences that are evident between groups using parallel forms of a measure (Gregorich 2006). This approach evaluates the presence of four common forms of measurement bias: (1) factor structure (Does the scale have the same number of underlying factors across groups?) (2) factor loadings (Does the scale convey the same meaning across groups?) (3) item intercepts (Do groups differ in their response profiles? For instance, does one group exhibit a more acquiescent response style?), and (4) item residual variances (is item true score measurement more reliable in one group than another?). Unless it has been demonstrated that parallel forms of a scale are free of any of the first three forms of measurement bias, one may question the validity of conclusions about substantive group differences. Furthermore, in order to validly compare group means using item composite scores, rather than latent means as is the more common practice, it is necessary to also establish invariance in item residual variances (Gregorich 2006). 1.2 Study Aims The aim of this study is to evaluate the psychometric equivalence of the adult PWI and two alternate versions of the PWI for adolescents (PWI-SC and PWI-SCb, respectively) using tests of measurement invariance as described above. Although both forms of the PWI have an eighth domain (school satisfaction for adolescents, and satisfaction with religion/spirituality for adults), these two domains are, at face value, conceptually distinct and therefore unsuitable to be tested for measurement invariance. Accordingly, only the seven domains which overlap across groups will be tested for invariance.

2 Methodology 2.1 Group 1: An Adolescent Sample Using ‘Sad’–‘Happy’ Scale (PWI-SC) Participants were 678 students recruited from a high-school in metropolitan Melbourne in 2005. The sample was 19% female and ages ranged from 11 to 18 (M = 14.28 years,

123

916

A. J. Tomyn et al.

SD = 1.75 years). Participants responded using an 11-point, end-defined scale (0 = very sad; 10 = very happy). 2.2 Group 2: An Adolescent Sample Using ‘Dissatisfied’–‘Satisfied’ Scale (PWI-SCb) This sample is the same as that used by Tomyn and Cummins (2011a). These were 351 students, recruited in 2005 and 2006 from several high-schools in the Melbourne metropolitan region. The sample was 59.3% female and ages ranged from 12 to 20 (M = 15.70 years, SD = 1.75 years). Participants responded using an 11-point, enddefined scale (0 = very dissatisfied; 5 = neutral; 10 = very satisfied). 2.3 Group 3: An Adult Sample Using ‘Satisfied’ Scale (PWI-A) The sample comprised a nationally representative sample of 1965 participants forming Survey 22 of the Australian Unity Wellbeing Index (Cummins et al. 2009). The sample was 51.1% female and ages ranged from 18 to 92 years (M = 51.81 years, SD = 17.05 years). Participants responded using an 11-point, end-defined scale (0 = completely dissatisfied; 5 = neutral; 10 = completely satisfied). 2.4 Measures Two versions of the personal wellbeing index-school children (PWI-SC; Cummins and Lau 2005; and PWI-SCb, Tomyn and Cummins 2011a) and a single version of the personal wellbeing index-adult (PWI-A; IWG 2006) were employed. The PWI-SC, as outlined in the user manual (Cummins and Lau 2005) and described previously, comprises seven items (domains) of happiness which are selected to represent the first level deconstruction of happiness with ‘life as a whole’ (Global life happiness; GLH). The students in Group 1 responded to domains using the adjectives ‘sad’ and ‘happy’ while those in Group 2 responded using the adjective ‘satisfied’ (see Appendix 1). Cronbach’s a for both versions of the PWI-SC were adequate (Cronbach’s a = 0.82); while Cronbach’s a for the PWI-A was 0.78. Means, standard deviations, and correlations between variables for group data are presented in Appendix 2. 2.5 Procedure Group 1 and 2 participants completed the PWI-SC in their regular classroom under conditions of privacy and informed consent in 2005 and 2006. The adult data were collected via a telephone interview conducted over the period 7th September to 19th September, 2009. Interviewers asked to speak to the person in the house who had the most recent birthday, was at least 18 years of age, and who was fluent in English. The group means, standard deviations and within-group correlations are presented in Appendix 2. Comparative Australian adult normative ranges can be found in Cummins et al. (2011). 2.6 Data Analytic Strategy Two separate multi-group confirmatory factor analyses were conducted in AMOS version 17 to test: (1) whether the two adolescent versions of the PWI function differently from each

123


917

other, and (2) whether these adolescent versions differ from the adult form of the PWI. These analyses proceeded as follows. First, the domains of satisfaction are modeled as a unidimensional construct (consistent with prior research and theory). This is done separately for each group (baseline model 1 for PWI-SC; baseline model 2 for PWI-SCb and baseline model 3 for adults). Adequacy of model fit is examined using the following criteria: Comparative Fit Index (CFI C 0.95 for good fit), Root Mean Square Error of Approximation (RMSEA; RMSEA \ 0.05 for good fit, B0.08 for adequate fit), and Standardized Root Mean Square Residual (SRMR; SRMR\0.05 for good fit) (Byrne 2010; Hu and Bentler 1999). Once adequate model fit is established for each group separately, multiple-group Means and Covariance Structure modeling is used to assess measurement invariance between groups (model 4 compares the two adolescent samples, whereas model 5 compares all three samples). Four increasingly stringent invariance assumptions (models 4a–4d, and 5a–5d, respectively) are tested in sequence, starting with the least restrictive model. Models 4a and 5a (configural invariance) require that items load onto the same factors across data sets, but allows item parameters (factor loadings, residual variances, and intercepts), factor variances and latent means to vary across groups. In models 4b and 5b (weak invariance), equality constraints across groups are applied to factor loadings and model fit is re-evaluated. Evidence of adequate fit for this model ensures that a given factor has the same meaning across groups (Gregorich 2006). Strong invariance (models 4c and 5c) involves constraining item intercepts to equality across groups to evaluate potential for systematic bias in responses from one group to another. If the assumptions of strong invariance hold, then an additional equality constraint is placed on residual variances (models 4d and 5d—strict invariance). This last step ensures that group differences obtained from comparisons of item composite scores (i.e., summing and averaging across individual items) can be attributed to substantive differences on the construct and are not due to differences in proportion of error variance in item-level scores. Measurement invariance is statistically evaluated by calculating differences in fit indices (typically, Dv2, DCFI, etc.) between reference and comparison models. The target model is typically compared against a less restrictive model (e.g., comparing model 1d versus model 1c). As v2 is sensitive to sample size and also to minor departures from normality (DiStefano and Hess 2005; Kline 2005), some researchers advocate the use of practical changes in model fit, using one of several comparative fit measures (e.g., CFI or TLI) as these indices have been shown to be relatively unaffected by sample size within the context of tests of measurement invariance (Cheung and Rensvold 2002). The present study used DCFI [ 0.01 to indicate practical change in fit from one model to the next, as recommended by Cheung and Rensvold (2002). The practice of constraining a factor loading to unity—while necessary to allow for calculation of scores on the latent variable(s) in a model—can lead to inaccurate parameter estimates of measurement invariance when the chosen factor loading, in fact, does vary across groups (Cheung and Rensvold 1999). As such, we used Cheung and Rensvold’s (1999) factor-ratio test to establish a suitable factor loading to constrain. This allows for an uncontaminated assessment of measurement invariance.

3 Results 3.1 Data Cleaning and Preparation All cases were examined for response sets. This is deemed to occur when a respondent consistently scores at the scale minimum (0) or maximum (10) for all seven of the PWI

123

918

A. J. Tomyn et al.

domains. These cases are considered unreliable and subsequently removed prior to the main analyses. 12 response sets were evident on the PWI-SC and of these, 11 were scale maximum scores and one was a scale minimum score. No other cases were removed. Across each of the three data sets, SPSS frequency output revealed that the frequency of missing data for all variables across respective data sets was less than 5%. Given that the patterns of missing data appeared random, missing values were replaced by regression, as recommended by Tabachnick and Fidell (2007). Examination of z-scores revealed univariate outliers on domain satisfaction variables. Comparison of mean scores on these variables with corresponding means trimmed at the upper and lower 5% showed that none of these outliers significantly influenced mean scores on key variables. Consequently, these univariate outliers were retained within further analyses (Pallant 2001). Absolute skew and kurtosis values were within the acceptable ranges of \2.0 and \7.0 respectively (Curran et al. 1996), thus demonstrating that domain scores were normally distributed. However, subsequent tests revealed evidence of multivariate kurtosis; Mardia’s coefficient = 26.76, p \ 0.05. Accordingly, a bias-corrected bootstrapping approach (on 2,000 samples of the original data set) was used to correct for multivariate non-normality, and to allow for more accurate estimation of model fit statistics in measurement invariance tests (Byrne 2010). The means and standard deviations for each group, presented in Appendix 2, can be referenced to normative ranges found in Cummins et al. (2011: Table A2.21) While the mean PWI for adults (76.19 points) and Group 2 adolescents (73.73) lie within the normal range (73.68–76.72), the mean for Group 1 (77.18) lies 0.46 points above the range. In terms of the domains, all lie within the normative ranges with the exception of one domain in Group 1 (‘Community’ is ?6.48 points above the range, and three in Group 2 (‘Community’ ? 1.88, ‘Health’—1.70, and ‘Achieving’—4.01). It is evident that ‘Community’ appears consistently above the adult level in both adolescent samples. In terms of the within-group inter-domain correlations, all lie in the range 0.2–0.5 and are significant at p \ 0.001. There are no obvious differences in either the levels or patterns of these correlations between the samples. In summary, with the possible exception of Community, the basic profile of the three groups looks similar. 3.2 Test for Measurement Invariance As shown in Table 1, the baseline model provides adequate fit of the data across all three groups, suggesting that each version of the PWI measures a uni-dimensional construct. Although the v2 values were all significant and RMSEA values were marginally acceptable, this likely reflects moderate to large sample sizes rather than model misspecification given that: (1) other fit indices (i.e., CFI and SRMR) were all at acceptable levels, and (2) inspection of modification indices failed to reveal any pathways that would substantially improve overall model function.

Table 1 Baseline model fit summary Model

v2

df

v2/df

CFI

SRMR

RMSEA

1. PWI-SC

69.67**

14

4.98

0.96

0.03

0.08

2. PWI-SCb

32.65*

14

2.33

0.98

0.04

0.06

139.43**

14

9.96

0.96

0.03

0.08

3. PWI-A

* p \ 0.01, ** p \ 0.001

123


919

Table 2 Evaluations of measurement invariance Model

v2

df

v2/df

Ref model

DCFI

CFI

SRMR

RMSEA

Invariance comparison 1 (PWI-SC vs PWI-SCb) 4a. Configural

102.32**

28

3.65

–

–

0.97

0.04

0.05

4b. Weak

103.87**

34

3.06

1a

0.01

0.96

0.04

0.04

4c. Strong

154.07**

40

3.85

1b

0.01

0.95

0.04

0.05

4d. Strict

190.41**

46

4.14

1c

0.01

0.94

0.04

0.05

Invariance comparison 2 (PWI-SC vs PWI-SCb vs PWI-A) 5a. Configural

241.75**

42

5.76

–

–

0.96

0.04

0.04

5b. Weak

243.78**

54

4.51

1a

\0.01

0.96

0.04

0.04

5c. Strong

307.58**

66

4.66

1b

0.01

0.95

0.04

0.04

5d. Strict

358.97**

78

4.60

1c

\0.01

0.95

0.05

0.04

* p \ 0.01, ** p \ 0.001

In Table 2, the top portion represents the evaluation of measurement invariance across the parallel forms of the PWI-SC, whereas the bottom portion represents a comparison of all three versions of the PWI (PWI-A, PWI-SC, and PWI-SCb). It is clear from Table 2 that placing increasingly stringent invariance conditions on the data failed to lead to a substantial reduction in model fit for model 4 or 5. At no point do these models exhibit a change in CFI that exceeds the (practically significant) cut-off of 0.01. This suggests that the parallel forms of the adolescent PWI perform comparably to each other (model 4), and that they also function equivalently to the adult version of the PWI (model 5, lower portion of Table 2). Furthermore, at each step, the fit statistics for these two models suggest that the models provide an adequate representation of the data.

4 Discussion The aim of this study was to use tests of measurement invariance to evaluate the extent to which two adolescent versions of the Personal Wellbeing Index for children (PWI-SC and PWI-SCb) function equivalently to each other, and to determine whether these parallel forms exhibit comparable psychometric properties to the original adult version (PWI-A). In particular, we evaluated whether the three scales have the same factor structure (configural invariance) and underlying meaning (weak invariance), whether they exhibit similar levels of response bias (strong invariance), and are subject to similar levels of error in measurement (strict invariance). By establishing these forms of invariance, defensible quantitative comparisons of SWB between adolescent and adult populations can be made using the PWI (Gregorich 2006). Our findings support the cross-group equivalence of parallel versions of the PWI and also provide further evidence that the PWI is a uni-dimensional measure of subjective wellbeing (IWG 2006). Despite differences in item wording, age and gender differences across samples, our analyses confirmed that a single factor model of the PWI items provided an adequate fit of the data. Furthermore, the imposition of increasingly stringent cross-group equivalence rules (for factor loadings, item intercepts and item error variances) failed to worsen this model fit. This pattern of measurement equivalence was found when comparing the two adolescent versions of the PWI against each other, and also when comparing these versions against the original, adult version of the PWI.

123

920

A. J. Tomyn et al.

These findings can be used to inform the differences noted by Tomyn and Cummins (2011a) when comparing the relative contributions of the PWI domains to GLS in adolescent and adult samples. For adolescents, these authors found that satisfaction with ‘Safety’ contributed unique variance while the domains of ‘Relationships’ and ‘Community connectedness’ did not. This pattern is the reverse of that found within adult samples, where the domains of ‘Relationships’ and ‘Community connectedness’ usually make a unique contribution, while ‘Safety’ does not. It seems likely that these earlier results reflect substantive differences in the importance of the various domain contributions to GLS, as opposed to measurement artifacts. They support the idea that domains under threat (e.g., Relationships) may be counter-balanced by higher satisfaction/happiness in other domains (e.g., Safety), thereby maintaining overall SWB within the normal range. This phenomenon, termed ‘domain compensation’ by Best et al. (2000), has been proposed as one of the homeostatic devices. 4.1 Which Version of the PWI-SC is Superior? While the finding of measurement equivalence for the two adolescent versions means that either version can be used, we recommend using the original PWI-SC (Cummins and Lau 2005) with younger participants (e.g., aged 11–13 years), or when participants have poor vocabulary and/or cognitive ability. Although there is no direct empirical support offered for this recommendation in the present study, the PWI-SC incorporates the adjective ‘happy’ because it is argued to be more easily comprehended and is less abstract than the adjective ‘satisfied’. As evidence of this difference, the word ‘satisfied’ does not feature on the Oxford Wordlist (Oxford University Press 2008) as one of the 307 most frequently used words by children of primary school age (‘happy’ features as word number 130); or the Dale-Chall wordlist (Chall and Dale 1995), in which these authors describe words on this list as amongst the most elemental words in the English language. Also in support of this difference, ‘Time4Learning’Ó (2011) offers a ‘learning to read’ program that assists students from early to mid-secondary school develop reading skills. The word ‘happy’ features on the second grade word list, indicating this is a word that students should know and are expected to spell correctly. However, the word ‘satisfied’ does not feature on this word list, even up to the eighth grade, suggesting that this is not a word commonly used amongst students at this level. In summary, this study demonstrates the cross-group equivalence of adult and adolescent versions of the PWI-SC in an Australian sample using tests of measurement invariance. This measurement approach can be used for other purposes, such as to establish the equivalence of the PWI translated for different cultural groups. 4.2 Limitations A limitation of this study is that while the adult sample is normative, the adolescent samples are not. The convenience nature of the adolescent samples is reflected by the fact that over three-quarters of participants who completed the PWI-SC were male; whereas two-thirds of participants who completed the PWI-SCb were female. It was also found that the adolescent samples scored higher in the domain of Community. However, the findings of invariance likely indicate the robust nature of these results in the face of such demographic and domain differences. Additionally, it should be acknowledged that both the adolescent samples included children with a broad range of ages (11–18 years and 12–20 years respectively). Consequently, if

123


921

comparisons were made between younger and older adolescents, differences may emerge, for example, due to differences in cognitive function from early to late adolescence. The implication here is that future research should conduct tests of measurement invariance for children of different age groups, for example, children in early, mid and late adolescence, to determine whether responses vary as a function of cognitive developmental milestones throughout adolescence. Finally, although our adolescent samples covered a broad and representative age range (11–20 years), our findings cannot speak to the adequacy of SWB measurement in preadolescent samples. Such populations are challenging due to lower levels of reading ability and comprehension (Cummins and Lau 2004). The personal wellbeing index—pre school (PWI-PS; Cummins and Lau 2004) may prove more suitable for children younger than 11 years due to the use of less complicated item wording and incorporation of a pre-testing protocol to determine whether, and to what level of complexity, respondents are able to use the scale. 4.3 Summary and Conclusions This study is the first to examine the cross-group equivalence of the adult and school children versions of the PWI using multiple-group Confirmatory Factor Analysis. Our finding of equivalence suggests that the meaning of the SWB construct is the same for adults and adolescents. This, in turn, supports the validity of quantitative comparisons between the PWI-A and both versions of the PWI for high-school age children.

Appendix 1 See Table 3. Table 3 Wording of PWI-A, PWI-SC and PWI-SCb Domain

PWI-A

PWI-SC

PWI-SCb

1. Standard of living

How satisfied are you with your standard of living?

How happy are you about the things that you have? Like the money you have and the things you own?

How satisfied are you about the things that you have? Like the money you have and the things you own?

2. Health

How satisfied are you with your health?

How happy are you with your health?

How satisfied are you with your health?

3. Achieving in life

How satisfied are you with what you are achieving in life?

How happy are you with the things that you want to be good at?

How satisfied are you with the things that you want to be good at?

4. Relationships

How satisfied are you with your personal relationships?

How happy are you about getting on with the people you know?

How satisfied are you about getting on with the people you know?

5. Safety

How satisfied are you with how safe you feel?

How happy are you with how safe you feel?

How satisfied are you with how safe you feel?

123

922

A. J. Tomyn et al.

Table 3 continued Domain

PWI-A

PWI-SC

PWI-SCb

6. Community connectedness

How satisfied are you with feeling part of your community?

How happy are you about doing things away from your home?

How satisfied are you about doing things away from your home?

7. Future security

How satisfied are you with your future security?

How happy are you about what may happen to you later on in your life?

How satisfied are you about what may happen to you later on in your life?

8. Religion/ Spirituality

How satisfied are you with your spirituality or religion?

N/A

N/A

9. School*

N/A

How happy are you with your school?

How satisfied are you with your school?

* The domain of ‘School’ will feature in the next revision (4th ed.) of the PWI-SC

Appendix 2 See Tables 4, 5 and 6. Table 4 Group 1 means, standard deviations and correlations between variables (N = 678) Variable

Mean

SD

1.

2.

3.

4.

5.

6.

7.

8.


79.04

18.71

–

2. Health

76.53

21.20

0.35

–


73.48

19.06

0.36

0.46

–

4. Relationships

79.19

17.54

0.41

0.41

0.46

–

5. Safety

80.58

17.52

0.47

0.38

0.40

0.43

6. Community connection

79.31

18.32

0.24

0.29

0.29

0.35

0.42

–

7. Future security

72.17

20.11

0.39

0.39

0.46

0.35

0.46

0.39

–

8. PWI

77.18

13.05

0.66

0.69

0.71

0.70

0.73

0.61

0.72

–

7.

8.

–

* All correlations significant at p \ 0.001

Table 5 Group 2 means, standard deviations and correlations between variables (N = 351) Variable

Mean

SD

1.


73.36

19.80

–

2. Health

71.85

21.27

0.44

–


67.92

19.30

0.41

0.54

–

4. Relationships

79.12

18.18

0.38

0.47

0.51

–

5. Safety

79.80

19.74

0.33

0.40

0.41

0.50


74.71

21.45

0.25

0.30

0.26

0.37

0.39

–

7. Future security

69.34

19.98

0.45

0.50

0.53

0.46

0.45

0.37

–

8. PWI

73.73

14.07

0.66

0.74

0.74

0.74

0.70

0.61

0.76


123

2.

3.

4.

5.

6.

–

–


923

Table 6 Group 3 means, standard deviations and correlations between variables (N = 1965) Variable

Mean

SD

1.

2.

3.

4.

5.

6.

7.


79.84

15.74

–

2. Health

74.21

20.08

0.41

–


74.02

18.51

0.48

0.38

–

4. Relationships

79.69

21.23

0.36

0.24

0.45

–

5. Safety

80.75

17.00

0.33

0.25

0.30

0.27


72.07

19.33

0.34

0.20

0.36

0.30

0.34

–

7. Future security

72.75

19.11

0.45

0.30

0.44

0.30

0.43

0.39

–

8. PWI

76.19

12.40

0.71

0.60

0.74

0.65

0.61

0.63

0.71

8.

–

–


References Best, C. J., Cummins, R. A., & Lo, S. K. (2000). The quality of rural and metropolitan life. Australian Journal of Psychology, 52(2), 69–74. Byrne, B. M. (2010). Structural equation modeling in AMOS: Basic concepts, applications, and programming (2nd ed.). New York: Routledge. Chall, J. S., & Dale, E. (1995). Readabiliy revisited: The new Dale-Chall readability formula. Cambridge: Harvard University Press. Cheung, G. W., & Rensvold, R. B. (1999). Testing factorial invariance across groups: A reconceptualization and proposed new method. Journal of Management, 25, 1–27. Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9, 233–255. Cummins, R. A. (1995). On the trail of the gold standard for subjective wellbeing. Social Indicators Research, 35, 179–200. Cummins, R. A. (2010). Subjective wellbeing, homeostatically protected mood and depression: A synthesis. Journal of Happiness Studies, 11, 1–17. Cummins, R. A., Collard, J., Woerner, J., Weinberg, M., Lorbergs, M., & Charini, P. (2009). Australian unity wellbeing index, survey 22: The Wellbeing of Australians—Who makes the decisions, health/ wealth control, financial advice, and handedness. Melbourne: Australian Centre on Quality of Life, School of Psychology, Deakin University. Retrieved May 12, 2010, from www.deakin.edu.au/research/ acqol/auwbi/survey-reports/survey-022-report-part-a.pdf. Cummins, R. A., & Lau, A. D. L. (2004). Personal wellbeing index: Pre-school (PWI-PS) (3rd ed). Melbourne: Deakin University. Retrieved February 9, 2011, from http://www.deakin.edu.au/research/ acqol/instruments/wellbeing-index/. Cummins, R. A., & Lau, A. D. L. (2005). Personal wellbeing index: School children (PWI-SC) (3rd ed). Melbourne: Deakin University. Retrieved January 17, 2010, from http://www.deakin.edu.au/research/ acqol/instruments/wellbeing_index.htm. Cummins, R. A., Lau, A. L. D., & Stokes, M. (2004). HRQOL and subjective well-being: noncomplementary forms of outcome measurement, Expert review of pharmacoeconomics & outcomes research, 4, 413–420 (Future Drugs, London, England). Cummins, R. A., & Nistico, H. (2002). Maintaining life satisfaction: The role of positive bias. Journal of Happiness Studies, 3, 37–69. Cummins, R. A., Woerner, J., Hartley-Clark, L., Perera, C., Gibson-Prosser, A., & Collard, J. (2011). Australian unity wellbeing index: Report 25.0—The Wellbeing of Australians—Relationships and the internet. Melbourne: Australian Centre on Quality of Life, School of Psychology, Deakin University. ISBN 978 1 74156 160 9 http://www.deakin.edu.au/research/acqol/index_wellbeing/index.htm. Cummins, R. A., Woerner, J., Weinberg, Perera, C., Gibson, A., & Collard, J. (2010). Australian unity wellbeing index, survey 23: The wellbeing of Australians—Life better/worse, children and neighbourhood. Melbourne, VIC: Australian Centre on Quality of Life, Deakin University. Retrieved March 12, 2010, from: http://www.deakin.edu.au/research/acqol/auwbi/survey-reports/survey-023-reportpart-a.pdf.

123

924

A. J. Tomyn et al.

Curran, P. J., West, G. W., & Finch, J. F. (1996). The robustness of test statistics to nonnormality and specification error in confirmatory factor analysis. Psychological Methods, 1, 16–29. Davern, M., Cummins, R. A., & Stokes, M. (2007). Subjective wellbeing as an affective-cognitive construct. Journal of Happiness Studies, 8, 429–449. DiStefano, C., & Hess, B. (2005). Using confirmatory factor analysis for construct validation: An empirical review. Journal of Psycho educational Assessment, 23, 225–241. Gregorich, S. E. (2006). Do self-report instruments allow meaningful comparisons across diverse population groups? Testing measurement invariance using the confirmatory factor analysis framework. Medical Care, 44, S78–S90. Headey, B., & Wearing, A. (1989). Personality, life events, and subjective wellbeing: toward a dynamic equilibrium model. Journal of Personality and Social Psychology, 57(4), 731–739. Headey, B., & Wearing, A. (1992). Understanding happiness: A theory of subjective wellbeing. Melbourne: Longman Cheshire. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55. International Wellbeing Group. (2006). Personal wellbeing index—Adult. Melbourne: Australian Centre on Quality of Life, Deakin University. Retrieved March 13, 2010, from: http://www.deakin.edu.au/ research/acqol/instruments/wellbeing_index.htm. International Wellbeing Group. (2011). http://www.deakin.edu.au/research/acqol/iwbg/index.php. Kline, R. B. (2005). Principles and practice of structural equation modeling. New York: Guilford Press. Oxford University Press. (2008). Oxford Wordlist. Australia and New Zealand. Pallant, J. (2001). SPSS survival manual. Sydney: Allen & Unwin. Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). MA: Allyn & Bacon. Time4Learning. (2011). Second grade spelling help. Retrieved September 21, 2011, from http://www. time4learning.com/SpellingWords/2nd-grade-spelling-words.shtml. Tomyn, A. J., & Cummins, R. A. (2011a). The subjective wellbeing of high-school students: Validating the personal wellbeing index-school children. Social Indicators Research, 101, 405–418. Tomyn, A. J., & Cummins, R. A. (2011b). Subjective wellbeing and homeostatically protected mood: Theory validation with adolescents. Journal of Happiness Studies, 12(5), 897–914. doi:10.1007/ s10902-010-9235-5.

123