Measuring and understanding eudaimonic well-being: a bifactor exploratory structural equation modelling approach

(1)

A bifactor exploratory structural equation modelling

approach

A Cromhout

orcid.org/0000-0002-0008-8212

Thesis

accepted in fulfillment of the requirements for the degree

Doctor of Philosophy in Health Sciences with Positive

Psychology at the North-West University

Promoter:

Prof L Schutte

Co-promoter:

Prof MP Wissing

Graduation:

Student number:

December 2020

11792833

(2)

Acknowledgements

My sincerest gratitude is expressed to my Lord and Saviour, Jesus Christ, for making this dream possible and for carrying me every step along the way. My gratitude is also expressed to the following persons who supported me in this endeavour:

1. To my promotor, Prof. Lusilda Schutte. Your dedicated guidance, encouragement, and approach to learning contributed to my academic growth. It has been a pleasure and a privilege to be mentored by you.

2. To my co-promotor, Prof. Marié Wissing. As always it has been a privilege to work with you. Your encouragement, expert advice, strong work ethic, and knowledge never cease to inspire.

3. To the language editor, Dr. Marietjie Nelson. Thank you for the hours spent to enhance the quality of my thesis.

4. To my family. Your never-ending love, support, and encouragement cannot be expressed in words. I am privileged to have you in my life. To my mother, who dreamed this dream with me, I wish you were still here to share this moment with me.

5. To the North-West University, South Africa. Thank you for the financial support in the form of a doctoral degree scholarship and a bursary from the Faculty of Health Sciences.

6. To the National Research Foundation (NRF) of South Africa. Thank you for funding the FORT research project and this PhD-study. This work is based on the research supported in part by the National Research Foundation of South Africa (Grant Number: 106050, 123266, 120060, 121948). The grantholders acknowledge that opinions, findings and conclusions or recommendations expressed in any publication generated by the NRF-supported research are those of the authors, and that the NRF accepts no liability whatsoever in this regard.

(4)

Summary

Measuring and understanding eudaimonic well-being: A bifactor exploratory structural equation modelling approach

Keywords: eudaimonic well-being, basic psychological needs theory, relational health, peer

relationships, community relationships, psychometric properties, factorial validity, reliability, measurement invariance, bifactor exploratory structural equation modelling

The overall aim of this study was to measure and understand eudaimonic well-being (EWB) by exploring the psychometric properties and measurement invariance of selected measures of EWB using recently developed statistical analytical techniques. Across three manuscripts recently developed statistical analytical techniques, that address some of the limitations inherent to traditional statistical techniques, were applied to three measures of eudaimonic well-being that operationalise three well-being theories respectively.

In Manuscript 1, CFA, bifactor CFA, ESEM, and bifactor ESEM were applied to the English (n = 326 ), Afrikaans (n = 478), and Setswana (n = 260) versions of the 21-item Basic Psychological Needs Scale (BPNS, Gagné, 2003) in three nonprobability South African student samples. For all language versions of the BPNS the three-factor bifactor ESEM model displayed superior fit compared to the other models that were fitted, but the fit was still inadequate for the BPNS-English and BPNS-Setswana and several items had to be removed to obtain models with adequate fit. A single item was removed from the bifactor ESEM model for the BPNS-Afrikaans that initially displayed close to adequate fit. Problematic scale items were removed because of problematic item formulation and scale construction. The general psychological need satisfaction

(5)

factor showed sufficient reliability scores for all language versions of the BPNS, and only the Competence and Relatedness specific factors of the BPNS-Afrikaans showed sufficient reliability indicators. Partial metric invariance of a reduced three-factor bifactor ESEM model was supported in the English and Afrikaans groups. Although the BPNS-Afrikaans showed potential for use, alternative measures of basic psychological needs should be considered for the current English and Setswana groups. The universal application of basic psychological needs is questioned in a South African context.

In Manuscript, 2 CFA, bifactor CFA, ESEM, and bifactor ESEM were applied to the English (n = 326, Sample 1), Afrikaans (n = 478, Sample 2) and Setswana (n = 260, Sample 3) versions of the Questionnaire for Eudaimonic Well-being (QEWB, Waterman et al., 2010) in three nonprobability South African student groups, and the English version of the QEWB in a nonprobability South African adult sample (n = 261, Sample 4). One-, three-, and four-factor structures were fitted to the data. For student Samples 1-3 the one-factor model displayed poor fit. Although the four-factor structure displayed slightly improved fit to the three-factor structure, the latter was selected for parsimony. The three-factor bifactor ESEM model fitted Samples 1-3, although item 9 had to be removed for Sample 3. The general factor, and some specific factors, obtained sufficient reliability scores for the student samples, pointing towards the potential for use of the QEWB in the current samples. Configural invariance of the QEWB among the student samples was supported. For adult Sample 4, all models tested yielded poor model fit. The results were likely influenced by the developmental phase of the participants. Future research should explore the manifestation and measurement of eudaimonic well-being in different developmental stages.

(6)

In Manuscript 3, CFA, bifactor CFA, ESEM, and bifactor ESEM were applied to the Relational Health Indices – Peer (RHIP) and Community (RHIC) Scales (Liang et al., 2002) in two nonprobability South African adult samples. Sample 1 (n = 286) completed the RHIP and RHIC in English and Sample 2 (n = 400) completed the RHIC in Setswana. A reduced three-factor bithree-factor ESEM model displayed adequate fit in all cases. Problematic scale items were removed from the RHIP and RHIC based on statistical grounds and substantive reasons that included items that did not capture the intended meaning of the target factor. For Samples 1 and 2 the reliability scores were sufficient for the general peer and community relational health factors respectively. For Sample 1 only some specific factors obtained sufficient reliability scores for the RHIP and RHIC, but none of the specific factors obtained sufficient reliability scores for Sample 2. The RHIP and RHIC show potential for use, provided that a general peer or community relational health is interpreted, and subscale scores are used with caution. Future research may develop a Construct Problem Solving factor and adapt the Authenticity factor to include more positively worded items that are representative of this factor.

Overall, the results pointed towards the multidimensionality of eudaimonic well-being as measured by the BPNS, QEWB, RHIP, and RHIC, while supporting the existence of a global eudaimonic well-being factor that coexist with some specific factors. The results emphasized the importance of proper conceptualisation and operationalisation of constructs when scales are adapted, translated, or developed. Cultural or contextual variables are important considerations in this regard.

(7)

Preface

This thesis is submitted in fulfilment of the requirements for the Doctor of Philosophy in Health Sciences with Positive Psychology, where the thesis accounts for all 360 credits. The thesis is submitted in article format as described by the General Academic Rules (2020) of the North-West University, specifically rules A.5.3.1.1, 5.3.2, 5.4.1, 5.10.1, 5.10.3, 5.10.4, and 5.10.5.

Three manuscripts have been prepared. Manuscript 1 (Section 2) was submitted to Current Psychology. Manuscript 2 (Section 3) will be submitted to Psychological Reports, and

Manuscript 3 (Section 4) to the Journal of Social and Personal Relationships. Manuscripts have been prepared in accordance with the author guidelines of the respective journals. The remainder of the thesis, that includes the Introduction (Section 1) and the Summary, Conclusions, and Recommendations (Section 5), have been prepared in accordance with the 7th edition of the Publication Manual of the American Psychological Association. For the purpose of this thesis page numbering is consecutive although each manuscript started from page 1 for submission to the respective journals. Where unsubmitted manuscripts exceed the prescribed word count, they will be abbreviated before submission to the journal. Standard margin width was maintained for uniformity. For ease of reference, tables and figures were inserted where they are first referred to in the manuscripts. Supplementary tables were inserted after the reference list of each

manuscript.

The letter of permission from the co-authors to submit these manuscripts for the Ph.D. degree is attached.

(8)

Letter of Permission

Permission is hereby granted by the co-authors that the following manuscripts may be submitted by Amanda Cromhout for the purpose of obtaining a Doctor of Philosophy in Health Sciences with Positive Psychology:

1. Factor structure and measurement invariance of the Basic Psychological Needs Scale in three South African samples: A bifactor exploratory structural equation modelling approach

2. The Questionnaire for Eudaimonic Well-being in student and adult South African samples: A bifactor exploratory structural equation modelling approach

3. The factor structure of the Peer and Community Relational Health Indices: A bifactor exploratory structural equation modelling approach

Contributions of the authors: Marié P. Wissing (MPW) developed, obtained ethical approval, and acquired funding for the broad FORT3-project. The above listed three manuscripts form part of the FORT3-project. For each of the three manuscripts, Amanda Cromhout (AC), Lusilda Schutte (LS), and MPW designed and planned the specific study. LS and MPW were responsible for the data gathering and capturing. The statistical analyses were performed and interpreted by AC and LS. AC drafted each manuscript, incorporated the recommendations from the co-authors,

prepared the final manuscript for submission, and served/will serve as the corresponding author during the submission process. LS and MPW continuously provided feedback regarding the intellectual content of the manuscripts. AC was responsible for the technical and language editing of the manuscripts. The authors read and approved the final manuscripts.

Prof L Schutte (Promotor)

(9)

(10)

Section 1: Introduction

Researchers and philosophers alike spent many years on endeavours to understand the concept of well-being. This is evidenced by the many theories that try to explain well-being from different perspectives and/or philosophical groundings. While these theories greatly increased our knowledge about well-being and the constituents thereof, there is still much to be known.

One field that aims to understand and empirically explore psychosocial well-being is the field of Positive Psychology. Positive Psychology is described as “the study of the conditions and processes that contribute to the flourishing or optimal functioning of people, groups, and institutions” (Gable & Haidt, 2005, p. 104). Two overlapping, yet distinct, approaches to well-being are often discerned, namely hedonia and eudaimonia (Huta & Waterman, 2014; Ryan et al., 2008). Hedonic well-being became associated with subjective well-being, which is often linked with having higher levels of positive emotions, lower levels of negative emotions, and higher levels of life satisfaction (Diener, 1984; Diener et al., 1985; Diener et al., 2017; Waterman, 1993). Positive and negative affect represent the affective facet of subjective well-being, and life satisfaction represents the cognitive-judgmental facet thereof (Diener et al., 1985). Eudaimonic well-being, on the other hand, is less clearly defined. It is mostly associated with facets indicative of “functioning well” (Martela & Sheldon, 2019). For example, eudaimonic well-being is associated with psychological well-being which, according to Ryff (1989), include six characteristics, namely self-acceptance, personal growth, autonomy, positive relationships, environmental mastery, and purpose in life. It is also associated with personal expressiveness (Waterman, 1993), being fully functional and able to deal with the challenges of life (Ryff & Singer, 2008), and living the life that one was meant to live (Deci & Ryan, 2008). Martela and Sheldon (2019) found that eudaimonic well-being was, at the point of their study, conceptualised

(11)

by at least 63 constructs which were operationalised in at least 45 different measures of eudaimonic well-being, with meaning or purpose, autonomy, competence, relatedness, and engagement as the elements that are most often measured.

The distinction between hedonic and eudaimonic well-being has been criticised as unsupported by empirical evidence (Kashdan et al., 2008), mainly because of high correlations found between hedonic and eudaimonic well-being (cf. Disabato et al., 2016; Kim et al., 2016). However, other studies supported the distinction between hedonic and eudaimonic well-being (Johansloo, 2016). Ryan et al. (2008) argued that hedonic and eudaimonic well-being are interrelated since living the good/well-lived life (eudaimonic well-being) may result in hedonic satisfactions, while pleasure and positive affect may facilitate positive human functioning.

Whether hedonic and eudaimonic well-being represent a single construct, two distinct constructs, or overlap in some respects, researchers seem in agreement that both are needed to experience well-being (Kim et al., 2017).

Although both hedonic and eudaimonic experiences contribute to well-being, Sheldon et al. (2019) found that pursuits to increase one’s eudaimonic well-being have greater pay-offs than attempts to increase one’s hedonic well-being over longer periods of time. Ryan et al. (2008) argued that a focus on hedonic outcomes alone will not result in lasting individual or collective well-being, without the addition of eudaimonic outcomes that may result in a more meaningful life and longer lasting hedonic outcomes.

Considering the importance of experiencing eudaimonic well-being, it is essential that researchers and practitioners properly understand the various facets of eudaimonic well-being, and should be able to measure these in a valid, reliable, and contextually relevant way. To this end, theory-based, developed and psychometrically sound measures of eudaimonic

(12)

well-being are necessary. Additionally, we need proper statistical techniques that can be applied to explore the psychometric properties of measures. If statistical techniques are based on

assumptions that do not hold in practice, the conclusions drawn from these analyses may be of limited use. Use of alternative improved statistical techniques are then advised. This study will focus on measures of eudaimonic well-being, and within this stream of thought focus specifically on basic psychological needs theory, a subtheory of self-determination theory (Deci & Ryan, 2000), the relational-cultural theory (Miller, 1976), as well as Waterman et al’s. (2010)

conceptualisation of eudaimonic well-being. Each of these theories as well as the selected scales operationalising them, will be presented in the paragraphs following below.

Theories and Operationalisations of Eudaimonic Well-being

The Basic Psychological Needs Theory: A Subtheory of Self-determination Theory

Self-determination theory (SDT) is a theory that describes human motivation and personality development. The theory emphasises that inner resources are important for

personality development and self-regulation (Ryan & Deci, 2000; Ryan et al., 1997). It examines the inherent tendency to grow and the innate psychological needs that are the foundation of self-motivation and personality integration, and further examines the conditions that are necessary to promote these processes (Ryan & Deci, 2000).

Basic psychological needs theory (BPNT), a subtheory of SDT, postulates that

environments supportive of psychological need satisfaction will contribute to growth and optimal functioning, while ill-being will result when environments thwart psychological need satisfaction (Deci & Ryan, 2000; Olafsen et al., 2017; Rigby & Ryan, 2018). The theory proposes three innate basic psychological needs, namely autonomy which refers to the need to experience self-determination and to have a sense of personal choice, competence which refers to the need for

(13)

effective interaction with the environment, and to feel capable of executing tasks on different levels of difficulty, and relatedness which refers to the need to have close relationships with others, and to feel that one is taken care of and supported by others (Deci & Ryan, 2000; Olafsen et al., 2017; Rigby & Ryan 2018; Ryan et al., 2008).

According to BPNT the satisfaction of all three basic psychological needs is required to attain psychological well-being outcomes (e.g., growth, optimal functioning). When any one or more of these basic psychological needs are frustrated, negative psychological outcomes (e.g., nonoptimal functioning, compensatory behaviour patterns) will result (Deci & Ryan, 2000; Olafsen et al., 2017). Basic psychological needs are postulated to apply cross-culturally and the consequences associated with need satisfaction and need dissatisfaction are expected to replicate across cultures (Chen et al., 2015; Deci & Ryan, 2000). The Basic Psychological Needs Scale (Gagné, 2003) measures basic psychological need satisfaction and is introduced in the next paragraph.

The Basic Psychological Needs Scale (BPNS, Gagné, 2003). The 21-item Basic

Psychological Needs Scale (BPNS, Gagné, 2003) is based upon BPNT (Deci & Ryan, 2000) and measures basic psychological need satisfaction. Several studies found that the a priori three-factor structure (distinguishing Autonomy, Competence, and Relatedness) best fitted the data, but that model fit was still inadequate, and reduced models from which problematic scale items were removed, revealed more adequate model fit (cf. Cromhout et al., 2018; Johnston & Finney, 2010; Schutte et al., 2018; Sheldon & Hilpert, 2012). These studies used confirmatory factor analysis (CFA), which is based on strict assumptions of the independent cluster model (ICM) that may not hold in practice (Howard et al., 2018), to explore the factor structure of the BPNS. Furthermore, Schutte et al. (2018) found that the English and Afrikaans versions of the 21-item

(14)

BPNS were partially scalar invariant, but a different baseline model fitted the Setswana version of the measure, which placed possible question marks over the notion that the theory applies universally. The current study is the first to apply recently developed statistical techniques (discussed in the section “Recent Approaches to Factor Analysis: ESEM, Bifactor CFA, and Bifactor ESEM”) to the BPNS. The BPNS will be discussed in more detail in Section 2 of the thesis.

Eudaimonic Well-being as Conceptualised by Waterman et al. (2010)

According to Waterman et al. (2010), eudaimonic well-being can be conceptualised in terms of six interlinked categories that have strong connections with philosophy and psychology. Self-discovery is essential for self-actualisation and knowing one’s identity. Perceived

development of one’s best potential involves that one identifies and develops those facets that represent one’s best potential. A sense of purpose and meaning in life is about deciding to which personally meaningful objectives one’s talents and skills will be applied. Investment of

significant effort in pursuit of excellence refers to the tendency to invest more effort in those

activities that are deemed personally meaningful. Intense involvement in activities refers to how intensely one is involved in activities that are personally meaningful. Enjoyment of activities as personally expressive refers to one’s involvement in activities that are expressive of one’s self.

Objective eudaimonic well-being elements (i.e., behaviours related to eudaimonic goal pursuits), as well as subjective eudaimonic well-being elements (i.e., what people experience when they are committed to excellence in fulfilling their personal potential), are included in this conceptualisation of eudaimonic well-being (Waterman et al., 2010). Waterman et al. (2010) referred to these subjective experiences of eudaimonia as “feelings of personal expressiveness” (p. 42) and discerned it from subjective well-being (hedonia), indicating that the former is the

(15)

result of the pursuit of one’s life purpose and the development of one’s potential, while the latter represents a desired outcome in itself.

This conceptualisation of eudaimonic well-being is operationalised by the Questionnaire for Eudaimonic well-being (Waterman et al., 2010). The measure is introduced in the next paragraph.

The Questionnaire for Eudaimonic Well-being (Waterman et al., 2010). The Questionnaire for Eudaimonic Well-being (QEWB, Waterman et al., 2010) is based upon Waterman et al’s (2010) conceptualisation of eudaimonic well-being. Waterman et al. (2010) applied parcelling to data from two ethnically diverse American student samples and concluded that a unifactorial structure best fitted the scale. Schutte et al. (2013) argued that Waterman et al. (2010) did not test the assumption of unidimensionality within parcels and therefore questioned the use of parcelling. Applying confirmatory factor analysis (CFA) and exploratory factor analysis (EFA) to data from a multicultural South African student group, Schutte et al. (2013) found support for a multidimensional structure consisting of three factors (Sense of Purpose, Purposeful Personal Expressiveness, Effortful Engagement) or four factors (Sense of Purpose, Engagement in Rewarding Activities, Living from Beliefs, Effortful Engagement). The three-factor structure was suggested for parsimony.

While some recent studies supported the unidimensionality of the QEWB, others supported the multidimensionality of the QEWB. Areepattamannil and Hassim (2017) applied CFA and found support for a unidimensional factor structure in an Indian adolescent sample. Applying Rasch-analysis, Sotgiu et al. (2019) found that the Italian version of the QEWB had a unidimensional structure in an Italian adult sample. Applying bifactor exploratory structural equation modelling (bifactor ESEM) to the three- and four-factor solutions proposed by Schutte

(16)

et al. (2013), Fadda et al. (2017) found support for the multidimensionality of the Italian version of QEWB in an Italian student sample, as well as for the existence of a global factor. Klym-Guba and Karaś (2018) applied CFA, EFA, and exploratory structural equation modelling (ESEM) to data from young adult (the average age of their four samples ranged between 20 and 24 years) Polish samples who completed the Polish version of the QEWB and found support for a three-factor ESEM solution, where the three-factors found by Schutte et al. (2013) were distinguished. The current study is the first to apply recently developed statistical techniques (discussed in the section “Recent Approaches to Factor Analysis: ESEM, Bifactor CFA, and Bifactor ESEM”) to the English, Afrikaans, and Setswana versions of the QEWB in South Africa to student samples and the English version of the QEWB to older adult samples. The QEWB will be discussed in more detail in Section 3 of the thesis.

Relational-cultural Theory

Rooted in feminism, relational-culture theory (RCT, Miller, 1976) is a developmental theory that was originally developed to explain the relational experiences of women and marginalised minority groups, that were not appropriately represented by traditional

developmental theories (Miller & Striver, 1997). Today it is accepted that the theory also applies to the relational experiences of men, and culture is regarded as an important consideration for the understanding of relational experiences (Alvarez & Lazzarie, 2016; Jordan, 2017; Liang et al., 2007).

The theory holds that humans experience growth, happiness, and well-being to the extent that they are engaged in growth-fostering relationships (Jordan, 2008). Relational competence and the ability to form meaningful, growth-fostering relationships are viewed as essential developmental outcomes, and stand in contrast to traditional developmental theories that

(17)

emphasise developmental outcomes such as autonomy, separation, and self-sufficiency (Jordan, 2001). Notions that are central to RCT are that growth takes place through human connections throughout life, increased relational competence characterises psychological growth, mutuality is a sign of mature functioning, authenticity is vital for genuine engagement, and mutual empathy and mutual empowerment are essential for growth-fostering relationships (Jordan, 2000; Miller 1976).

According to RCT, growth-fostering relationships comprise four attributes (Jordan 2010; Miller & Striver, 1997). Authenticity refers to the capacity for genuine engagement with others and to be one’s real self in interpersonal relationships. Mutual engagement refers to one’s commitment and attunement to the relationship and the ability to sustain a sense of self while being receptive to the change experiences that are presented within the relationship.

Differences/conflict refers to the ability to manage differences/conflict in such a way that they

can be expressed, dealt with, and accepted. Zest/empowerment refers to a sense of personal strength and feeling inspired towards action (Jordan 1986, 1997, 2010; Miller & Striver, 1997).

Well-being outcomes such as having an increased sense of zest; clarity about the self, others, and the relationship; a sense of self-worth; increased capacity for creativity and

productivity; and a desire for more connection with others (Jordan, 2008, 2010; Miller, 1986) are associated with having growth-fostering relationships and are referred to as “the five good things” (Jordan, 2008, 2010; Miller, 1986). Even when disconnection, disappointment, and feelings of isolation and withdrawal are experienced in the context of growth-fostering relationships, parties may still experience growth as they deal with the disconnection (Jordan, 2008, 2010). The Relational Health Indices (Liang et al., 2002) measure relational health in various relational contexts. The scales are introduced next.

(18)

The Relational Health Indices (Liang et al., 2002). The Relational Health Indices (RHI, Liang et al., 2002) are based on the relational-cultural theory (RCT, Miller, 1976) and measure relational health in peer (RHIP), community (RHIC), and mentor (RHIM) contexts (Liang et al., 2002). Each scale can be used separately and is proposed to consist of three factors, namely Authenticity, Engagement, Zest/Empowerment. However, results from studies that explored the factor structure of the RHI were inconsistent. Liang et al. (2002) found support for the three-factor structure in a female sample (Liang et al., 2002), but in a sample that also included male participants, a unidimensional factor structure produced better fit for each scale (Liang et al., 2007). Based on a sample consisting of male and female participants, Frey et al. (2005) found that for the RHIP and RHIM a unidimensional structure fitted best, with a two-factor structure obtaining best fit for the RHIC. According to Lenz et al. (2016) a three-factor structure best fitted the Spanish version of the scales in a sample with male and female participants. In all these studies CFA was applied to determine the factor structure of the scales, except in the study of Frey et al. (2005) where principal component analysis with oblimin rotation was applied. The present study is the first to apply recently developed statistical techniques (ESEM, Bifactor CFA, and Bifactor ESEM) to the RHI. The RHIP and RHIC are considered in this study and will be discussed in more detail in Section 4 of the thesis.

The questionable model fit and the inconsistent results obtained with regard to the dimensionality when traditional approaches to factor analysis (EFA and CFA) were applied to the BPNS, QEWB, and the RHIP and RHIC, necessitates a further exploration of the

psychometric properties of the scales by applying recently developed approaches to factor analysis (bifactor CFA, ESEM, and bifactor ESEM), designed to overcome the limitations of the traditional approaches. In the following paragraph, the assessment of the validity and reliability

(19)

of psychological measures will be discussed briefly. Traditional and recently developed procedures to explore the factor structure of scales will get specific attention. Measurement invariance will also be discussed.

Assessment of the Validity, Reliability, and Measurement Invariance of Psychological Measures

With the growing interest in a better understanding of well-being and the constituents thereof, as well as in the promotion of well-being, the rigorous measurement of well-being constructs is vital. Good measuring instruments provide several benefits for research and

practice. For example, they enable researchers to test the theories and conceptualisations of well-being that underly the various measuring instruments (cf. Chen et al., 2015; Howard et al., 2018; Johansloo, 2016; Johansloo et al., 2016; Jovanović, 2015). In this regard, good measuring

instruments not only provide insight into the factor structures of measures of well-being, but they also contribute to the (further) development of theory in that it enables researchers to determine the extent of the applicability of these theories in different populations and contexts (cf. Chen et al., 2015; Eriksson & Boman, 2018; Guhn et al., 2018). Good measuring instruments further allow researchers and practitioners to accurately assess levels of well-being and to evaluate the efficacy of interventions (Foxcroft & Roodt, 2018). For this purpose, it is essential that evidence for the validity and reliability of measuring instruments, as well as their equivalence, are

investigated in different contexts and for different groups (De Kock & Foxcroft, 2018). In the next section, validity as a psychometric property of measuring instruments will be discussed.

Validity

The validity of an instrument refers to the degree to which the instrument measures what it sets out to measure (Roodt & De Kock, 2018a). Roodt & De Kock (2018a) discusses three

(20)

types of validation procedures, namely content description procedures, criterion-prediction procedures, and construct identification procedures. This study will focus on the construct identification procedures, and more specifically on factorial validity, which focuses on the extent to which an instrument operationalises the hypothesised underlying latent factors (Roodt & De Kock, 2018a).

For purposes of this study, factor analysis (using recently developed statistical techniques) was applied to determine the factorial validity of the measures of interest. The following paragraphs explain traditional and recent approaches to factor analysis.

Traditional Approaches to Factor Analysis: EFA and CFA. To determine the factor structure of an instrument factor analysis can be used. Factor analysis is used to investigate the interrelationships (covariation) among a set of observed variables (indicators) in order to obtain information on the underlying latent constructs (Byrne, 2012; Kline, 2013; Roodt & De Kock, 2018a). By identifying the common variance between sets of variables a large number of

variables can be reduced to a small number of factors (or dimensions). These factors describe the factorial structure of the scale and assist in identifying subscales (Byrne, 2012; Roodt & De Kock, 2018a). Thus, factor analysis determines how, and to what extent, the indicators are associated with the underlying latent variables (Byrne, 2012).

Two main approaches to factor analysis are commonly applied, namely exploratory factor analysis (EFA) and confirmatory factor analysis (CFA, Byrne, 2012). EFA is applied when the associations between the observed and latent variables are not known or are uncertain (Byrne, 2012). The measurement model is unrestricted, which means that there is not a specified

indicator-factor association (Kline, 2013). This unrestricted measurement model is unidentified, which means that there is not a unique set of statistical estimates for a particular model (Kline,

(21)

2013). Thus, each indicator (observed variable) is regressed on each factor (representing the latent variable; Kline, 2013) and thus all possible pattern coefficients (factor loadings) are calculated for each indicator (Kline, 2013).

In contrast, CFA is applied when there is some theoretical or empirical knowledge that proposes expected associations between the observed and latent variables (Byrne, 2012). The measurement model is restricted, which means that there is a specific indicator-factor association (Kline, 2013). This restricted model is identified, which means that there is a unique set of estimates for a particular model (Kline, 2013). However, unlike with EFA, the cross-loadings of indicators on nontarget factors are constrained to be zero for standard CFA, so that each indicator only regresses on the target factor (Howard et al., 2018). This constraint is applied because standard CFA is based upon the independent cluster model (ICM) which specifies that the cross-loadings of indicators on nontarget factors are assumed to be zero (Howard et al., 2018). This constraint may result in biased parameter estimates as two sources of construct-relevant multidimensionality are not accounted for (Morin, Arens, & Marsh, 2016).

The first source of construct-relevant multidimensionality relates to the nature of the indicators used to measure constructs (Howard et al., 2018; Morin, Arens, & Marsh, 2016). Indicators are rarely related to the target factor only and will mostly also have construct-relevant associations with nontarget factors if the constructs represented by the factors are conceptually-related or hierarchically-ordered (Howard et al., 2018). If these associations (factor loadings) are forced to be zero, potentially true influence of nontarget factors on the indicator may be ignored. This may result in reduced goodness-of-fit indices as sources of misspecification are concealed. It may also impact on the discriminant validity of the factors as biased estimates of factor correlations may generate artificial multicollinearity when these factors are used for prediction

(22)

(Howard et al., 2018). Exploratory structural equation modelling (ESEM) provides a possible solution to this limitation of CFA, and will be discussed in the next subsection.

A second source of construct-relevant multidimensionality relates to constructs that are hierarchically-ordered (Howard et al., 2018; Morin, Arens, & Marsh, 2016). Higher-order factor models hypothesise that multiple factors can combine to form one or more higher-order factors (Howard et al., 2018), where the first-order factors fully mediate the associations between the items and the higher-order factors (Morin, Arens, Tran, & Caci, 2016). This implies that the first-order factor reflects both the variance explained by the higher-first-order factor and the unique

variance explained by each first-order factor (Morin, Arens, Tran, & Caci, 2016). As an

alternative to hierarchical models, bifactor models allow for the direct estimation of the influence of the general factor on the indicators. ESEM and bifactor modelling will be introduced in the next section.

Recent Approaches to Factor Analysis: ESEM, Bifactor CFA, and Bifactor ESEM. More recent trends in statistical analyses provide solutions to the constraints inherent in ICM CFA. One of these involve the integration of EFA within the structural equation modelling (SEM) framework (Morin, Arens, & Marsh, 2016). In this regard, exploratory structural equation modelling (ESEM, Asparouhov & Muthén, 2009), allows for a model to be defined according to CFA specifications, while items are allowed to cross-load on nontarget factors similar to EFA (Howard et al., 2018). ESEM accounts for the construct-relevant multidimensionality which is caused by the associations that an indicator has with the nontarget factors, over and above the association it has with the target factor, thus addressing the first source of construct-relevant multidimensionality discussed above.

(23)

Although EFA/ESEM is often critiqued for the fact that the inclusion of cross-loadings may taint the meaning of the latent constructs (factors), Morin, Arens, Tran, and Caci (2016) indicate that it is rather the exclusion of cross-loadings that taints the meaning of the construct. Simulation studies indicated that the incorporation of cross-loadings in ESEM provides more accurate estimates of the true population values for factor correlations. If ICM CFA assumptions fit the underlying population model, the estimates will remain unbiased (Asparouhov & Muthén, 2009; Howard et al., 2018). In this regard Morin, Arens, and Marsh (2016) indicate that small cross-loadings of indicators on nontarget factors can be regarded as the influence of the nontarget factor on the construct-relevant part of the indicator, provided that these cross-loadings

correspond to theoretical expectations. When cross-loadings are unexplainable, or large (e.g., larger than target loadings), it may imply that items and the factors to which they were assigned should be reexamined (Morin, Arens, & Marsh, 2016).

The second source of construct-relevant multidimensionality, which is caused by the hierarchical ordering of constructs, is addressed by bifactor models. Bifactor models test if a unitary global factor, which directly influences all the indicators, coexists with the specific factors (Howard et al., 2018). Furthermore, the variance that is attributable to the specific factors can be separated from the variance that is attributable to the global/general factor in bifactor models. The direct relations between the indicators and the specific and global factors are estimated. All of the variance that is shared among all indicators are forced to be absorbed into the global factor, while the specific factor represents the variance that is shared among a specific subset of indicators (Howard et al., 2018; Morin, Arens, Tran, & Caci, 2016). In bifactor models, indicators are used to describe both their a priori specific factors, as well as the global general factor (Howard et al., 2018).

(24)

When constructs that are conceptually-related and hierarchically-ordered are measured, they could be represented by models that incorporate both cross-loadings and a general factor. When the cross-loadings are not modelled in bifactor CFA models the estimates of the global factor may be inflated, while the cross-loadings in EFA models may be inflated when the general factor is not modelled (Morin, Arens, & Marsh, 2016). Bifactor ESEM models allow for the incorporation of cross-loading and a general factor (Jennrich & Bentler, 2011).

ESEM, bifactor CFA, and bifactor ESEM models provide solutions to the limitations inherent in ICM CFA by allowing for the incorporation of cross-loadings and/or a global factor. Therefore, the use of these models may lead to improved factorial validity of measurement instruments.

Examples of Applications of ESEM, Bifactor CFA, and Bifactor ESEM in Well-being Assessment. The contribution of ESEM, bifactor CFA, and bifactor ESEM can be illustrated by previous studies that applied these analytical procedures to various well-being constructs. As mentioned earlier, the distinction between hedonic and eudaimonic well-being is often criticised as lacking empirical support due to high interfactor correlations (Disabato et al., 2016; Kashdan et al., 2008) in studies that applied CFA.

Johansloo (2016) applied CFA and ESEM to investigate whether hedonic and

eudaimonic well-being are separate constructs. Hedonic well-being was measured in terms of positive and negative affect, by using the positive and negative affect scales of Mroczek and Kalarz (1998), and life satisfaction, by using indicators of life satisfaction relating to the individual’s overall life, work, health, relationship with spouse/partner, and relationship with children. Eudaimonic well-being was measured in terms of psychological well-being, using Ryff’s (1989) psychological well-being scales, and social well-being, using Keyes’ (1998) social

(25)

well-being scale. Johansloo (2016) obtained substantially smaller interfactor correlations with ESEM (Mr = .47) than with CFA (Mr = .71), thereby concluding that, although hedonic and

eudaimonic well-being are correlated, they are largely independent constructs. These results were supported in another study by Johansloo et al. (2016) where they investigated the factor structure of mental well-being. The application of CFA and ESEM indicated that ESEM

provided superior model fit and substantially smaller factor correlations (Mr= .35) than the CFA model (Mr = .74). In this study hedonic well-being was measured in terms of positive and

negative affect, using items from the Spanish version (Echevarria & Páez, 1989) of the Bradburn Affect Balance Scale (Bradburn, 1969); domain satisfaction, using items that measure

satisfaction in the family, money and income, friends, job, and oneself-as-a-person domains; and general life satisfaction, using a single item measure of general life satisfaction. Eudaimonic well-being was measured in terms of psychological well-being, using items from Ryff’s (1989) psychological well-being scale, and social well-being, using items from Keyes’ (1998) social well-being scale.

Turning to bifactor modelling, Jovanović (2015) tested several CFA and bifactor CFA models to reevaluate the structure of subjective well-being (SWB). SWB was measured in terms of three specific factors, namely Satisfaction with Life (SWL; from the Satisfaction with Life Scale [SWLS], Diener et al., 1985), and Positive Affect (PA) and Negative Affect (NA, from the Positive Affect Negative Affect Scale; Watson et al., 1988). The three-factor bifactor CFA model, with a general SWB factor and three specific factors, namely SWL, PA, and NA, produced superior fit. Most of the items of the SWLS and PA loaded strongly on the general factor (≥ 0.4) with higher loadings on the general factor than the specific factors, while the items of the NA had relatively strong loadings on the general factor (-0.28 to -0.60, average λ =.40),

(26)

but with stronger loadings on the specific factor than the general factor. The results were indicative of a strong general SWB factor while at the same time suggesting that SWL, PA, and NA reflect specific factors of SWB, thereby providing support for the multidimensionality of SWB. These findings, together with the omega reliability scores, suggested that the scores of the SWLS and the PANAS should be interpreted with caution and that a general SWB factor should be taken into account (Jovanović, 2015).

Another point of critique regarding the dimensionality of well-being constructs centres around whether psychological need satisfaction and frustration are two distinct constructs as found by Costa et al. (2015). Applying bifactor ESEM to the Basic Psychological Need Satisfaction and Frustration Scale, Tóth-Király et al. (2018) investigated the dimensionality of need fulfillment and tested several alternative models, amongst others a six-factor (satisfaction x fulfillment: autonomy, competence, relatedness) bifactor ESEM model with one general factor (need satisfaction) and a six-factor (satisfaction x fulfillment: autonomy, competence,

relatedness) bifactor ESEM model with two general factors (need satisfaction, need frustration). Although both of these models displayed superior fit in comparison with the other models tested, the two general factors had high correlations and were weakly defined, indicating that they could have been combined into one general factor. A six-factor bifactor ESEM model with one general factor was considered superior, thereby indicating that need satisfaction and need dissatisfaction are not two distinct constructs, but opposite ends on the same continuum. Furthermore, some of the specific factors retained meaningful specificity, indicating that some of the specific factors can reliably be interpreted as independent factors that coexist with the general need satisfaction factor (Tóth-Király et al., 2018).

(27)

These studies illustrated that models that allow for the incorporation of cross-loadings and/or a general factor often resulted in improved model fit, and possibly a more accurate reflection of parameter estimates, thereby resulting in more accurate conclusions about the dimensionality of constructs and measures of well-being (Howard et al., 2018; Morin, Arens, & Marsh, 2016). These findings question the utility of ICM CFA to understand and measure the various facets of well-being, due to the restrictive assumptions underlying ICM CFA (Howard et al., 2018; Morin, Arens, & Marsh, 2016). The application of ESEM, bifactor CFA, and bifactor ESEM to the selected well-being measures in the current study will allow for comparison of these models with the CFA models and not only shed light on the utility of CFA with regard to the selected well-being measures in the selected contexts, but also provide insight into the constructs on substantive level. In addition to looking for evidence for the validity of a measuring instrument, evidence for its reliability must also be established. Reliability will be discussed in the next section.

Reliability

The degree to which an instrument is consistent in measuring what it claims to measure, is referred to as the reliability of the measure (Roodt & De Kock, 2018b). Various forms of reliability exist, for example, test-retest reliability, inter-scorer and intra-scorer reliability, alternate-form reliability, and internal consistency reliability (Roodt & De Kock, 2018b). This study will focus on internal consistency reliability, which refers to the degree of consistency of the responses across scale items (Kline, 2016).

Although alpha is commonly used to determine internal consistency reliability, its application to multidimensional measures may be problematic. Alpha is based upon the essentially tau-equivalent model, that assumes a constant true score variance across all scale

(28)

items (Raykov, 1997). This assumption is not problematic in unidimensional measures where there are perfect intercorrelations between true scores of scale items, but this assumption is violated in multidimensional measures where intercorrelations between the true scores are not perfect (Dunn et al., 2014; Soĉan, 2000). Should alpha be applied to multidimensional measures, reliability estimates may be biased and result in attenuation or inflation of reliability scores (see Dunn et al., 2014).

An alternative to alpha is omega (McDonald, 1999). Omega is based upon the congeneric model that allows for variation of means and variances of the true scores and error variances (Jöreskog, 1971). Omega is based on fewer assumptions that are more realistic than the assumptions underlying alpha and it is less likely that internal consistency estimates will be attenuated or inflated. When the assumptions that underlie the essentially tau-equivalent model are satisfied, omega functions at least as well as alpha (Zinbarg et al., 2005), while omega

outperforms alpha when these assumptions are violated (Dunn et al., 2014). This study will apply omega as indicator of internal reliability as calculated by Sánchez-Oliva et al. (2017).

Apart from using scores on well-being measures for specific groups, researchers and practitioners may also be interested in group comparison on the constructs. Before groups can be compared, measurement invariance needs to be established. Measurement invariance is

introduced in the next section.

Measurement Invariance

Researchers are often interested in comparing different groups (e.g., based on gender, culture, age) in relation to a certain construct. Before groups can be compared, it must first be established if the instrument used to measure the construct functions equivalently across the various groups. Measurement invariance is used to assess the equivalence of an instrument

(29)

across groups or across different measurement time points and indicates if the underlying

construct has the same meaning across the different groups or across time (Putnick & Bornstein, 2017). Measures may be nonequivalent for a variety of reasons, for example, the construct may not have the same meaning for the different groups, or the meaning may only partially overlap in different groups, the meaning of the construct could have changed over time, or the

characteristics of the measurement instrument or the scale administration conditions could be problematic (e.g., De Kock & Foxcroft, 2018; He & Van de Vijver, 2012).

When evaluating the measurement invariance of an instrument, the following types of measurement invariance are discerned and are hierarchically-ordered (Putnick & Bornstein, 2017). In configural invariance or invariance of model form the same measurement model is tested across the groups. The number of factors and the factor-indicator specifications are the same, but all the parameters are freely estimated for each sample. If the model does not fit the data, then the measure is noninvariant on all levels (Kline, 2013). Metric invariance or

equivalence of item loadings means that the unstandardised pattern coefficients of all indicators

are equal within the limits of sampling error across the groups (Kline, 2013), in other words that each indicator contributes to a similar degree to the relevant latent variables across the groups (Putnick & Bornstein, 2017). If a measure is found to be metric invariant, it can be concluded that the construct is expressed similarly across groups (Kline, 2013). If not all of the pattern coefficients are equal, but the majority are, the measure is considered to display partial metric invariance (Kline, 2013; Putnick & Bornstein, 2017). Scalar invariance or equivalence of item intercepts requires that item intercepts, as well as factor loadings, are constrained to be equal

across the groups (Putnick & Bornstein, 2017). When a measure is scalar invariant between groups it implies that people who have similar scores on the latent construct will also score

(30)

similarly on the items. When scalar invariance is established, group means can be compared (Milfont & Fisher, 2010). When some, but not all item intercepts are equal, the measure is considered to display partial scalar invariance (Putnick & Bornstein, 2017). An investigation of the measurement invariance of the different language versions of the selected well-being measures will provide insight into the extent to which the scales operate equivalently across the groups.

The South African Context

South Africa is a culture-rich and ethnically diverse country with 11 official languages (Statistics South Africa, 2019). According to the midyear population estimates for 2019 (Statistics South Africa, 2019), the South African population is classified as Black African (80.7%), Coloured (8.8%), Indian/Asian (2.6%), and White (7.9%). South Africa is a developing country, facing many socio-economic challenges such as high crime rates, poverty,

unemployment, and limited access to health care (e.g., Gordon et al., 2020; Masipa, 2018). In 2015, 55.5% of South Africans were considered poor as measured against the upper-bound poverty line (UBPL, R992.00 per person per month in 2015) and 25.2 % were considered to live in extreme poverty (Statistics South Africa, 2017). In the third quarter of 2019, the

unemployment rate was 29.1% (Statistics South Africa, 2019).

These factors impact negatively on the psychosocial well-being of South Africans resulting in high prevalence of mental disorders such as depression, anxiety, and substance abuse. Limited access to health care, lack of mental health care providers, stigmatisation of mental disorders, cultural bias regarding mental disorders, and poor help-seeking behaviours of affected persons are all hindrances to curbing mental disorders, especially in rural communities (cf. Carbonell et al., 2020; Gordon et al., 2020; Jensen et al., 2020). It is clear that intervention is

(31)

needed on multiple levels to address these challenges and to buffer against their impact. One step towards such intervention would be to increase the psychosocial well-being of South Africans by considering relevant theory to design interventions and to evaluate their effectiveness by

applying psychometrically sound measures. This would require a contextual approach where the possible influence of cultural and contextual differences on the psychometric properties of measurement tools are considered. Measurement of psychological constructs and, actually, the nature and manifestation of the underlying constructs are largely unexplored in African contexts (Nwoye, 2015). One step in building a contextually relevant science of well-being in South Africa that can help to address challenges regarding the well-being of South Africans, is to explore the psychometric properties of instruments of important well-being constructs.

Explorations of the validity of measures of well-being will not only facilitate the evaluation of levels of psychosocial well-being in South African samples by researchers and practitioners, but also contribute to a better understanding of how specific theories apply cross-culturally.

The Present Study

The overall aim of this study was to measure and understand eudaimonic well-being by exploring the psychometric properties and measurement invariance of selected measures of eudaimonic well-being using recently developed statistical analytical techniques. Specifically, recently developed statistical analytical procedures, that aim to overcome the limitations inherent to traditional approaches to factor analysis, will be applied. Specifically, CFA, bifactor CFA, ESEM, and bifactor ESEM will be applied to data on the 21-item Basic Psychological Needs Scale (Gagné, 2003; English, Afrikaans, and Setswana versions), the Questionnaire for

Eudaimonic Well-being (Waterman et al., 2010; English, Afrikaans, and Setswana versions) and the Relational Health Indices (Liang et al., 2002; English version of the Peer Scale and the

(32)

English and Setswana versions of the Community Scale), to investigate the scales’ psychometric properties. This study will further explore the measurement invariance between different groups for some of these measures. Findings will be reported for purposes of this thesis in three

manuscripts followed by an integrating concluding section. The first manuscript (Section 2) investigates the psychometric properties and measurement invariance of the BPNS (Gagné, 2003) in three South African student groups who completed English, Afrikaans, and Setswana versions of the scale, respectively, and was submitted to Current Psychology. The second manuscript (Section 3) investigates the psychometric properties and measurement invariance of the QEWB (Waterman et al., 2010) in three South African student groups who completed the scale in English, Afrikaans, and Setswana, respectively, and one multicultural South African adult group who completed the English version of the scale. The manuscript will be submitted to Psychological Reports. The third manuscript (Section 4) investigates the psychometric properties

of the RHIP (Liang et al., 2002; English version) and RHIC (Liang et al., 2002; English and Setswana version) in two South African samples. A multicultural adult sample fluent in English completed the RHIP and RHIC in English and a Setswana-speaking adult sample completed the RHIC in Setswana. The manuscript will be submitted to the Journal of Social and Personal Relationships. The final concluding section (Section 5) contains the summary, conclusions, and

recommendations that are based on the findings of the present study.

With this study it is intended to make a contribution to the understanding and conceptualisation of eudaimonic well-being on a theoretical level, for example by providing insight into the dimensionality of basic psychological need satisfaction, eudaimonic well-being (as conceptualised by Waterman et al., 2010), and relational health in community and peer contexts. Furthermore, the results will shed light on the applicability of the scales and their

(33)

underlying theories in various cultural contexts, including the underexplored African context. Additionally, the application of traditional and recently developed statistical analysis allows for a comparison with regard to which of the tested statistical models provide the best representation of the selected eudaimonic well-being constructs. On a practical level, researchers and

practitioners can be informed by evidence that shows support or lack of support for the validity of the measures in the groups used in this study. This may provide useful information regarding the potential for using these instruments to assess levels of psychosocial well-being and evaluate interventions in similar groups.

(34)

References

Alvarez, A. R., & Lazzari, M. M. (2016). Feminist mentoring and relational cultural theory: A case example and implications. Journal of Women and Social Work, 31(1), 41-54. https://doi.org/10.1177/0886109915612512

Areepattamannil, S., & Hashim, J. (2017). The questionnaire for eudaimonic well-being

(QEWB): Psychometric properties in a non-Western adolescent sample. Personality and Individual Differences, 117, 236-241. http://dx.doi.org/10.1016/j.paid.2017.06.018

Asparouhov, T., & Muthén, B. O. (2009). Exploratory structural equation modelling. Structural Equation Modelling, 16(3), 397-438. https://doi.org/10.1080/10705510903008204

Bradburn, N. M. (1969). The structure of psychological well-being. Aldine

Byrne, B. M. (2012). Structural equation modeling with Mplus: Basic concepts, applications, and programming. Routledge.

Carbonell, A., Navarro-Pérez, J.-J., & Mestre, V.-M. (2020). Challenges and barriers in mental healthcare systems and their impact on the family: A systematic integrative review. Health and Social Care in the Community. http://dx.doi.org/10.1111/hsc.12968

Chen, B., Vansteenkiste, M., Beyers, W., Boone, L., Deci, E. L., Van der Kaap-Deeder, J., Duriez, B., Lens, W., Matos, L., Mouratidis, A., Ryan, R. M., Sheldon., K. M., Soenens, B., Van Petegem, S., & Verstuyf, J. (2015). Basic psychological need satisfaction, need frustration, and need strength across four cultures. Motivation and Emotion, 39(2), 216-236. https://doi.org/10.1007/s11031-014-9450-1

Costa, S., Ntoumanis, N., & Bartholomew, K. J. (2015). Predicting the brighter and darker sides of interpersonal relationships: Does psychological need thwarting matter? Motivation and Emotion, 39(1), 11-24. https://doi.org/10.1007/s11031-014-9427-0

(35)

Cromhout, A., Schutte, L., & Wissing, M. P. (2018). Validation of the Basic Psychological Needs Scale in a South African student group. South African Journal of Psychology, 48(4), 501-513. https://doi.org/10.1177/0081246317728267

Deci, E. L., & Ryan, R. M. (2000). The “what” and “why” of goal pursuits: Human needs and the self-determination of behaviour. Psychological Inquiry, 11(4), 227-268.

https://doi.org/10.1207/S15327965PLI1104_01

Deci, E. L., & Ryan, R. M. (2008). Hedonia, eudaimonia, and well-being: An introduction. Journal of Happiness Studies, 9(1), 1-11. https://doi.org/10.1007/s10902-006-9018-1

De Kock, F., & Foxcroft, C. (2018). Cross-cultural test adaptation, translation and tests in multiple languages. In C. Foxcroft, & G. Roodt (Eds.)., Introduction to psychological assessment in the South African context (5th ed., pp. 100-122). Oxford University Press. Diener, E. (1984). Subjective well-being. Psychological Bulletin, 95(3), 542-575.

http://dx.doi.org/10.1037/0033-2909.95.3.542

Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The Satisfaction with Life Scale. Journal of Personality Assessment, 49(1), 71-75.

http://dx.doi.org/10.1207/s15327752jpa4901_13

Diener, E., Heintzelman, S. J., Kushlev, K., Tay, L., Wirtz, D., Lutes, L. D., & Oishi, S. (2017). Findings all psychologists should know from the new science on subjective

well-being. Canadian Psychology, 58(2), 87–104. https://doi.org/10.1037/cap0000063

Disabato, D. J., Goodman, F. R., Kashdan, T. B., Short, J. R., & Jarden, A. (2016). Different types of well-being? A cross-cultural examination of hedonic and eudaimonic well-being. Psychological Assessment, 28(5), 471-482. http://dx.doi.org/10.1037/pas0000209

(36)

Dunn, T. J., Baguley, T., & Brunsden, V. (2014). From alpha to omega: A practical solution to the problem of internal consistency estimation. British Journal of Psychology, 105(3), 399-412. http://dx.doi.org/10.1111/bjop.12046

Echevarria, A., & Páez, D. (1989). Emociones: Perspectivas psicosociales [Emotions: A psychosocial perspective]. Fundamentos

Eriksson, M. & Boman, E. (2018). Short is beautiful: Dimensionality and measurement invariance in two length of the Basic Psychological Needs at Work Scale. Frontiers in Psychology, 9, 965-977. http://dx.doi.org/10.3389/fpsyg.2018.00965

Fadda, D., Scalas, L. F., Meleddu, M., & Morin, A. J. S. (2017). A bifactor ESEM representation of the Questionnaire for Eudaimonic Well-being. Personality and Individual Differences, 116, 216-222. http://dx.doi.org/10.1016/j.paid.2017.04.062

Foxcroft, C., & Roodt, G. (2018). An overview of assessment: Definition and scope. In C. Foxcroft, & G. Roodt (Eds.), Introduction to psychological assessment in the South African context (5th ed., pp. 3-11). Oxford University Press.

Frey, L. L., Beesly, D., & Newman, J. L. (2005). The Relational Health Indices: Reanalysis of a measurement of relational quality. Measurement and Evaluation in Counselling and Development, 38(3), 153-163. https://doi.org/10.1080/07481756.2005.11909776

Gable, S. L., & Haidt, J. (2005). What and why is Positive Psychology? Review of General Psychology, 9(2), 103-110.

Gagné, M. (2003). The role of autonomy support and autonomy orientation in prosocial behavior engagement. Motivation and Emotion, 27(3), 199-223.

(37)

Gordon, T., Booysen, F., & Mbonigaba, J. (2020). Socio-economic inequalities in the multiple dimensions of access to health care: The case of South Africa. BMC Public Health, 20(1), 1-13. https://doi.org/10.1186/s12889-020-8368-7

Guhn, M., Ark, T. K., Emerson, S. D., Schonert-Reichl, K. A., & Gadermann, A. M. (2018). The Satisfaction with Life Scale adapted for children: Measurement invariance across gender and over time. Psychological Assessment, 30(9), 1261-1266.

http://dx.doi.org/10.1037/pas0000598

He, J., & Van de Vijver, F. (2012). Bias and equivalence in cross-cultural research. Online Readings in Psychology and Culture, 2(2), 1-19. https://doi.org/10.9707/2307-0919.1111

Howard, J. L., Gagné, M., Morin, A. J. S., & Forest, J. (2018). Using bifactor exploratory structural equation modeling to test for a continuum structure of motivation. Journal of Management, 44(7), 2638-2664. https://doi.org/10.1177/0149206316645653

Huta, V., & Waterman, A. S. (2014). Eudaimonia and its distinction from hedonia: Developing a classification and terminology for understanding conceptual and operational definitions. Journal of Happiness Studies, 15(6), 1425-1456.

https://doi.org/10.1007/s10902-013-9485-0

Jennrich, R. I., & Bentler, P. M. (2011). Exploratory bi-factor analysis. Psychometrica, 76(4), 537-549. https://doi.org/10.1007/S11336-011-9218-4

Jensen, E. J., Wieling, E. & Mendenhall, T. (2020). A phenomenological study of clinicians’ perspectives on barriers to rural mental health care. Journal of Rural Mental Health, 44(1), 51-61. http://dx.doi.org/10.1037/rmh0000125

(38)

Johansloo, M. (2016). Revisiting the empirical distinction between hedonic and eudaimonic aspects of well-being using exploratory structural equation modeling. Journal of Happiness Studies, 17, 2023-2036. https://doi.org/10.1007/s10902-015-9683-z

Johansloo, M., Bobowik, M., & Basabane, N. (2016). Factor structure of mental well-being: Contributions of exploratory structural equation modeling. Personality and Individual Differences, 102, 107-110. http://dx.doi.org/10.1016/j.paid.2016.06.060

Johnston, M. M., & Finney, S. J. (2010). Measuring basic needs satisfaction: Evaluating previous research and conducting new psychometric evaluations of the Basic Needs Satisfaction in General Scale. Contemporary Educational Psychology, 35(4), 280-296.

https://doi.org/10.1016/j.cedpsych.2010.04.003

Jordan, J. V. (1986). The meaning of mutuality. Work in Progress, 23. Wellesley College, Stone Center for Developmental Services and Studies.

Jordan J. V. (1997). The relational model is a source of empowerment for women. In M. R. Walsh (Ed.), Women, men, and gender: Ongoing debates (pp. 373-382). Yale. Jordan, J. V. (2000). The role of empathy in relational-cultural therapy. Journal of Clinical

Psychology, 56(8), 1005-1016.

Jordan, J. V. (2001). A relational-cultural model: Healing through mutual empathy. Bulletin of the Menninger Clinic, 65(1), 92-103.

http://dx.doi.org.nwulib.nwu.ac.za/10.1521/bumc.65.1.92.18707

Jordan, J. V. (2008). Recent developments in relational-cultural theory. Women and Therapy, 31(2-4), 1-4. https://doi.org/10.1080/02703140802145540

(39)

Jordan, J. V. (2017). Relational-cultural theory: The power of connection to transform our lives. Journal of Humanistic Counselling, 56(3), 228-243.

https://doi.org/10.1002/johc.12055Jöreskog, K. G. (1971). Statistical analysis of sets of congeneric tests. Psychometrika, 36(1), 109–133. https://doi.org/10.1007/BF02291393

Jovanović, V. (2015). A bifactor model of subjective well-being: A re-examination of the structure of subjective well-being. Personality and Individual Differences, 87, 45-49. http://dx.doi.org/10.1016/j.paid.2015.07.026

Kashdan, T. B., Biswas-Diener, R., & King, L. A. (2008). Reconsidering happiness: The costs of distinguishing between hedonics and Eudaimonia. The Journal of Positive Psychology, 3(4), 219–233. http://dx.doi.org/10.1080/17439760802303044

Keyes, C. L. M. (1998). Social well-being. Social Psychology Quarterly, 61(2), 121–140.

Kim, K., Lehning, A. J., & Sacco, P. (2016). Assessing the factor structure of well-being in older adults: findings from the National Health and Aging Trends Study. Aging and Mental Health, 20(8), 814-822. http://dx.doi.org/10.1080/13607863.2015.1037245

Kim, J., Kim, Y., & Kim, D. (2017). Improving well-being through hedonic, eudaimonic, and social needs fulfilment in sport media consumption. Sport Management Review, 20(3), 309-321. http://dx.doi.org/10.1016/j.smr.2016.10.001

Kline, R. (2013). Exploratory and confirmatory factor analysis. In Y. Petscher, C.

Schatschneider, & D. L. Compton (Eds). Applied quantitative analysis in education and the social sciences (pp. 171-207). Routledge.

Kline, R. B. (2016). Data preparation and psychometrics review. Principles and practice of structural equation modeling (4th ed., pp. 64-96). The Guildford Press.

(40)

Klym-Guba, M., & Karaś, D. (2018). Polish version of the Questionnaire for Eudaimonic Well-being – three factors rather than one. Health Psychology Report, 6(3), 273-283.

https://doi.org/10.5114/hpr.2018.75684

Lenz, A. S. (2016). Relational-cultural theory: Fostering the growth of a paradigm through empirical research. Journal of Counseling and Development, 94(4), 415-428. http://dx.doi.org/10.1002/jcad.12100

Lenz, A. S., Balkin, R. S., Soler, I. G., & Martinez, P. (2016). Development and evaluation of a Spanish-language version of the Relational Health Indices. Psychological Assessment, 28(5), e62-e69. http://dx.doi.org/10.1037/pas0000170

Liang, B., Tracy, A., Taylor, C. A., Williams, L. M., Jordan, J. V., & Miller, J. B. (2002). The Relational Health Indices: A study of women’s relationships. Psychology of Women Quarterly, 26(1), 25-35.

Liang, B., Tracy, A., Glenn, C., Burns, S. M., & Ting, D. (2007). The Relational Health Indices: Conforming factor structure for use with men. The Australian Community Psychologist, 19(2), 35-52.

Martela, F., & Sheldon, K. M. (2019). Clarifying the concept of well-being: Psychological need satisfaction as the common core connecting eudaimonic and subjective well-being. Review of General Psychology, 23(4), 458-474.

http://dx.doi.org/10.1177/1089268019880886

Masipa, T. (2018). South Africa’s transition to democracy and democratic consolidation: A reflection on socio-economic challenges. Journal of Public Affairs, 18(4), 1-6. http://dx.doi.org/10.1002/pa.1713