Author | Date | No. of items | Sample size | Sample | Analysis method | Summary of results |
---|---|---|---|---|---|---|
Ryff, CD. | 1989 | 120 | 321 | Community volunteers; 3 age groups: young adults (mean age = 19.3 years, s.d. 1.6; midlife (49.9, 9.4) and older adults (75.0; 7.1). F = 60%. | Correlational analysis | Inter-factor correlations of a priori 6F model .32–.76; (highest: E with S = .76; S with P = .72; P with G = .72; P with E .66); internal consistency coefficients 0.86–0.93. |
Ryff, CD & Keyes, CL. | 1995 | 18 | 1,108 | Midlife in the United States (MIDUS). (Mean age = 45.6, 14.8); F = 59%. | LISREL 7.2 PRELIS WLS estimation continuous | Tested 6F model (BIC -167.6 df = 120), 6F with 2nd order (BIC -166.0, df = 129) and unidimensional model (-38.2, df = 135). Inter-factor correlations 0.30–0.85. (Highest E with S .85) |
Clark, PJ, Marshall VW, Ryff CD, Wheaton, B. | 2001 | 18 | 4,960 | Canadian Study of Health & Aging (mean age = 76) | EQS ML estimation (continuous) | Tested 1-6F models. Found low internal consistency & reliability for all models, some low factor loadings and a large number of cross-loadings. A priori 6F model (CFI = 0.77). Preferred solution – 6F model with 4 cross-loadings |
Kafka GJ & Kozma A. | 2002 | 120 | 277 | Canadian students (mean age = 21.3, 3.8) F = 67%. | Principal components analysis with varimax rotation. | Unrestricted model extracted 15 factors; a model restricted to 6F did not correspond to a-priori Ryff dimensions*. |
Van Dierendonck, D. | 2004 | 84 54 18 # | 233 420 | Dutch students (mean age = 22, 6); F = 67% Dutch professionals (mean age = 36, 8); F = 31%. | CFA – LISREL 8.5 ML estimation | A-priori 6F with 2nd order returned lower AIC than other models (e.g. 1F, 2F negative and positive, 5F & 6F) but CFI unacceptably low = 0.65 (84-item), 0.73 (54-item); 0.88 (18-item). Factorial validity only acceptable for 18-item version, but internal consistency was below acceptable limits for 84, 54 & 18-item. |
 |  | 54 | 420 | Dutch professionals (mean age = 36, 8); F = 31%. |  |  |
Cheng, ST, Chan, AC. | 2005 | 24 | 1,259 | Chinese aged 18–86 (mean age 44.7, 16.6); F = 82%, volunteers at hospitals in Hong Kong | CFA – LISREL (8.52)/PRELIS with ML estimation. | 6F model (CFI = 0.93); 1F (CFI = 0.88) and 6F + 2nd order (CFI 0.92). Initial CFA on 18-item scale with poorer model fit and low internal consistency than their revised 24-item version. |
Springer, KW, Hauser, RM. | 2006 | 42 12 18 18 | 6,282 6,038 2,731 9,240 | Wisconsin Longitudinal Survey MIDUS (25–74 yrs) National Survey of Families and Households NSFH II | CFA – LISREL/PRELIS Estimator = WLS (Polychoric correlations) Adjustment for correlated negative method factor and adjacent questions | Tested a range of models (including 1F, 6F + 2nd order). 6F a-priori model with correlated negative method factor, correlated errors for adjacent questions and 3 error correlations had lowest BIC; although model fit criteria was poor. High inter-factor correlations for all dimensions (.72–.97) Highest E with S; P with S. |
 |  | 18 | 2,731 | MIDUS (25–74 yrs) |  |  |
 |  | 18 | 9,240 | National Survey of Families and Households NSFH II |  |  |
Abbott, RA, Ploubdis, GB, Huppert, FA, Kuh, D, Wadsworth, MJ, Croudace, TJ. | 2006 | 42 | 1,179 | National Survey of Health & Development (UK) Women age 52. | CFA Mplus (3.1) rWLS estimator. Negative & positive method factors (uncorrelated) | Tested a range of models (including random allocation unidimensional, 6F + 2nd order). Method factors (orthogonal to the constructs and each other) for questions reflecting positive and negative item content improved model fit. Preferred model included a single 2nd order factor, loaded by 4 first-order factors (E,G,P,S), two method factors, and 2 distinct first-order factors (A,R) |