Study 1: Descriptive data and frequencies were calculated using SPSS soft-ware version 18.0 and scores for each of PWB scales are presented as means and standard deviation (SD). Items within each of the Ryff's PWB scales were summed to provide total scores and these were used in the confirmatory factor analysis and structural equation modeling. Cronbach's alpha coefficient was computed to assess internal consistency of the PWB scales.
To determine the relationships among gender, perceived health status, family re-lationship, and PWB, a Multiple Indicators Multiple Causes (MIMIC) structural equation model was performed using Mplus software version 5.1 (Muthén &
Muthén, 2008). The basic MIMIC model consists of two parts: A measurement model and a regression model. A MIMIC model is used here because it can test the factor structures of the PWB scales and Family Relationships using confirmatory fac-tor analysis (CFA) while simultaneously examining the regression of PWB on gen-der, perceived health and Family Relationships. CFA with the Maximum likelihood estimation was used as all the PWB scales and Family Relationships scores loaded onto one (forced) factor and no rotation was used. No case was excluded from analy-ses because of violations of multivariate normality and linearity.
Our MIMIC model allowed simultaneous estimation of the relation of two latent factors (PWB and Family Relationships) to gender and perceived health. It also al-lowed evaluation of indicator variables which are the six subscales from the Ryff PWB scales: autonomy, environmental mastery, personal growth, positive relations with others, purpose in life, and self-acceptance. It also incorporated the correlation between the factors of perceived health status and Family Relationships, while con-trolling for gender (Ríos-Bedoya, Pomerleau, Neuman, & Pomerleau, 2009). The cri-terion for an acceptable factor loading is at least 0.3 (Shevlin & Miles, 1998).
Cronbach’s alpha values equal to or greater than 0.70 were considered satisfactory (Jenkinson, Coulter, & Bruster, 2002).
Indices of model fit manifest how the proposed model may be consistent with data. Those used for refinement of the model and stopping criteria in this paper, from Kline (2005), are the relative/normed chi-square (χ2/df), comparative fit index (CFI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR). The chi-square value is the traditional measure for evaluat-ing overall model fit and, ‘assesses the magnitude of discrepancy between the sample and fitted covariances matrices’ (Hu & Bentler, 1999). The relative/normed chi-square (χ2/df) adjusts for the impact of sample size because the chi-chi-square statistic lacks power when the sample size is small. The CFI is a relative fit index that com-pares the χ2 value of the model to that of the null model, taking into account sample size and performing well even when sample size is small (Tabachnick & Fidell, 2007). The RMSEA tells us how well the model, with unknown but optimally chosen parameter estimates, would fit the population’s covariance matrix. The SRMR is the square root of the difference between the residuals of the sample covariance matrix
and the hypothesized covariance model. The cutoff values for refinement of the model and stopping criteria were (1) χ2/df < 2 (Tabachnick & Fidell, 2007), (2) CFI
≥ 0.95 (Hu & Bentler, 1999), (3) RMSEA < 0.07 (Steiger, 2007), and (4) SRMR <
0.05 (Bernabé et al., 2009).
Principle axis factor analysis with a varimax rotation was undertaken to evalu-ate the covariance of items and to identify derived factors of the CHI. Item loadings of >0.4 were considered adequate (Peterson, 2000) and retained. Cronbach’s alpha was calculated as the internal consistency reliability on retained factors and items of the CHI. A priori criterion of 0.7 was used to assess evidence of reliability (Jenkinson et al., 2002).
Canonical correlation analysis was used to examine the concurrent validity of the CHI with the PWB scales. Canonical correlation is a method of modeling the relationship between two sets of variables and to find linear combinations between two sets of multi-dimensional variables (Stewart & Love, 1968). The maximum number of canonical variates that can be extracted from the sets of variables equals the number of variables in the smallest set of variables. The software of R version 2.15.0 was utilized for canonical correlation analysis.
Study 2: Descriptive data and frequencies were calculated using SPSS soft-ware version 18.0 and scores for the CHI scale, sense of coherence, family relationships are presented as means and standard deviation (SD). Data were there-fore analyzed using Pearson’s product-moment correlation and path analysis with structural equation modeling method to detect the pathway on family relationships, sense of coherence, happiness and perceived health values.
To determine the relationships among family relationships, sense of coherence, perceived health status, and happiness, a path analysis was performed using Mplus software version 5.1 (Muthén & Muthén, 2008). Path analysis is an extension of the regression model, used to test the fit of the correlation matrix against two or more causal models. Path analysis allows the simultaneous modeling of several related re-gression relationships. A path analysis is used here because a variable can be a de-pendent variable in one relationship and an indede-pendent variable in another (Muthén, Muthén, 2008).
The path model tested was as follows: family relationships was hypothesized to influence sense of coherence and happiness status; sense of coherence would me-diate the influence of family relationships on the status of happiness; the happiness were hypothesized to influence perceived health status directly. Our path analysis model allowed simultaneous estimation of the relation of happiness and perceived health status. It also allowed to test the predictive and a mediating role of sense of coherence in happiness status (Pajares & Miller, 1994). Moreover, indicators of model fit were consistent with the structural equation modeling with study 1. The cutoff values for refinement of the model and stopping criteria were (1) χ2/df < 2 (Tabachnick & Fidell, 2007), (2) CFI ≥ 0.95 (Hu & Bentler, 1999), (3) RMSEA <
0.07 (Steiger, 2007), and (4) SRMR < 0.05 (Kline, 2005).
Study 3: To determine the process of broaden and build on happiness and per-ceived health status, the transcription of the whole interview were created using voice recorder. After transcribing, analyzing involves re-reading the interviewer tran-scripts to identify themes from respondents' answers. Using the topic and questions of the interview format to organize the analysis, in essence synthesizing the answers
to the questions the study proposed. Arrival at data saturation the content of analysis, the process of data collection were ended. The rule of data saturation was depending on the enough information to include to be able to carry out similar research them-selves. Otherwise, triangulation was used to verify the validity of the informations.