Brazilian Validation of Centrality of Religiosity Scale (CRS-10BR and CRS-5BR)

: The centrality of religiosity scale (CRS), available in three versions (with 5, 10 and 15 items), is a measuring instrument that identiﬁes the central importance of religiosity in the psychological construction and in the behavior of an individual. According to the literature, ﬁve components together express the centrality of religion in life: Public practice, private practice, ideological, intellectual, and religious experience. These components are the ground on which religious constructs are formed and activated. For the validation of the scale in the Brazilian cultural context, two versions were tested (CRS-10BR and CRS-5BR) with data collected from a general population (N = 687). Exploratory Factor Analysis (N = 334) resulted in a ﬁve-factor solution congruent to CRS-10BR. Conﬁrmatory Factor Analysis (N = 353) demonstrated that a ﬁve-factor solution (Intellect, Ideology, Private Practice, Public Practice and Religious Experience) indicated better ﬁt indexes than the single-factor solution of ﬁve items (CRS-5BR). Thus, CRS-10BR is recommended to capture CRS full construct. However, the CRS-5BR version can be considered suitable for use in the Brazilian population when the context is demanding simpler and faster data collection.


Introduction
The religious dimension of human subjectivity has multifaceted expressions. This is exhibited by the enormous variety of forms of religious expression that are shaped in society, stem from individual and collective experiences and are identified as a religious experience. Religiosity, while occupying a central place in the subjectivity of the individual, impacts decisions and predicts human behavior, whether this centrality is intrinsic or extrinsic. Religious beliefs impact the way people think, make choices, and behave (Allport and Ross 1967).
In Brazil, a country of approximately 200 million inhabitants, the importance of religion and spirituality/religiosity in both the individual and social domain is evident as indicated by the last religious census conducted by the Brazilian Institute of Geography and Statistics (IBGE). While Romain Catholicism, a predominant religion since colonization, appears to be declining progressively in all regions of the country, from 73.6% in the 2000 census to 64.6% in 2010, other religious groups have been gaining prominence, mainly Evangelicals, which jumped from 15.4% to 22.2% (IBGE 2010). Furthermore, the census shows that the Non-Religious group represent 8.03%. Among that 8.03%, about 4% declare to be atheists, and 0.8% are agnostics (IBGE 2010).

Theoretical Background of the Centrality of Religiosity Scale
The centrality of religiosity scale was created by Stefan Huber (2003Huber ( , 2004Huber ( , 2009 and is an instrument developed to measure the "centrality, importance or salience of religious meanings in personality" (Huber and Huber 2012, p. 711). It is presented in six versions. The versions of the present validation (CRS-10 and CRS-5) focus on the predominantly monotheistic Brazilian religious context, with an Abrahamic tradition, in which the majority is Christian. The original scale, in its different versions, reflects an interreligious pattern, including the CRS-15 with 15 items, the CRS-10 with 10 items and the shorter version CRS-5, with five items. The other three versions  were developed later in order to embrace new forms of spirituality in the Western context, in which meditation and mystical experiences are of great relevance (Huber and Huber 2012).
The centrality of religiosity scale was constructed on an interdisciplinary theoretical basis, at the interface between Theology/Studies of Religion, Sociology and Psychology, and distinguishes five core-dimensions of religion: Intellectual, ideological, private practice, public practice, and of religious experience; and three degrees of religiosity: Highly Religious, religious and Non-Religious.
Based on Huber and Huber (2012) work of validation, the five core-dimensions can be characterized as follows: Intellectual Dimension refers to the social expectation that religious people-regardless of religious affiliation or any denominational bias-have religious knowledge, interest, hermeneutical skills, styles of thought and interpretation that provides them with explanations on transcendence, religion and religiosity. The Ideological dimension comprehends that religious individuals have beliefs, convictions and standards of plausibility in the existence of a transcendent reality, being a fundamental assumption in order to construct the concepts and dogmas for most religious traditions. The Public Practice dimension refers to patterns of action in which, the religious individual with a sense of belonging to religious groups, participates and shares collective practices in certain activities and rituals. Similarly, the Private Practice dimension refers to the social expectation that religious individuals practice religious activities and rituals, but in the private space. Prayers and meditation are some examples. Through these practices, the basic patterns of action and personal style of devotion to transcendence and spirituality are expressed. The last dimension, Religious Experience, refers to the religious individual's experience of feeling connected to an ultimate reality, an experience of being one with something divine, or feeling part of something greater that affects them emotionally.
Huber made use of the central dimensions of religiosity from Charles Glock (1962Glock ( , 1973) studies in the field of Sociology of Religion, and adapted them, from the perspective of Personality Psychology, especially the concepts on intrinsic and extrinsic religiosity by Allport and Ross (1967), and the concepts on psychology of the personal religious construct-system by Kelly (1955). Theology, or Studies of Religion, contribute to the aspects related to the religious contents expressed in the five dimensions, such as concepts of God, theological concepts, religions quest, etc. (Table 1). Table 1. Items and versions (5 and 10) of the centrality of religiosity scale (CRS). * General contents (GC) and specific contents (SC). Adapted from Huber (2009).

Socio-Psychological Dimensions and Theological Contents * Items Versions
Intellect ( It is important to emphasize that both the intrinsic and extrinsic expression of religiosity can occupy a central place in the individual's life in order to impact the personal religious construct-system that guides their behavior. For Allport and Ross (1967), intrinsic religiosity is constant, devout and internalized. Extrinsic religiosity is casual, irregular and, in general, religion is used in an instrumental and utilitarian way. Although these "ideal types of religiosity" were constructed by Allport and Ross in a context that was different from the current Brazilian scenario, these concepts still offer contributions, because of the types of religiosity they describe. The expression of a religiosity that is intrinsic and well-integrated in the subjectivity, as well as a utilitarian and pragmatic expression of the extrinsic religiosity, can occupy centrality in the subject's life, if they are considered Highly Religious.

Centrality of the Personally Constructed System
The five dimensions described (Intellectual, Ideological, Public Practice, Private Practice and Religious Experience) can be seen as representative of the totality of religious life, and in this sense, the measure of their frequency and intensity expresses whether or not religiosity is central to the individual's life. The centrality of religiosity is observed by the sum of the scores. The total result is the sum of the results of the five factors. The sum of the indexes allows the categorization of the personal religious construct-system of individuals and/or groups into three levels: Highly Religious, Religious or Non-Religious.
For Highly Religious individuals, the religious system occupies a central position, meaning that the religious content and beliefs exert a strong influence on the psychological system.
As for the category "Religious", this refers to individuals whom: . . . the personal religious system holds merely a subordinate position within the personality's cognitive and emotional architecture. Though religious content can indeed be found in the person's life-horizon, it cannot be expected to have a clearly determinant effect on experience and behavior. Religion plays more of a background role. (Huber 2009, p. 21) Individuals in this category rarely use religious beliefs and values to construct meaning for their experiences. The third category raised by the scale characterizes the Non-Religious. In this group, religious contents do not appear, or very rarely appear in the personal religious construct-system of these individuals. "The presence of religiosity remains so weak that one cannot speak of the personality having an independent religious system-or if so, only for comparative purposes" (Huber 2009, p. 21).

Method
In order to access validation of the centrality of religiosity scale proposed by Huber and Huber (2012) in the Brazilian cultural context, data were collected from a general population using CRS-10BR (which also enables to test the CRS-5BR version of the scale) and sociodemographic questions. Therefore, the aim was to verify not only the dimensionality and reliability of both versions (CRS-10BR and CRS-5BR), but also comparing their results in terms of individuals' categorization of the centrality of religiosity across sociodemographic characteristics.
The first step of this process was the translation and the adaptation of its questions for the use in the Brazilian context, followed by separate Exploratory Factor Analysis and Confirmatory Factor Analysis to attest the dimensionality and reliability of both versions (CRS-10BR and CRS-5BR). As both versions demonstrated good reliability and mostly acceptable fit indexes, we performed agreement categorizations tests using Huber and Huber (2012) proposal in the extent of its CRS Score.

Translation and Adaptation Procedures
The centrality of religiosity scale is being validated in the versions CRS-5 (with five items) and CRS-10 (with 10 items). The process of translation, adaptation and validation for use in the Brazilian context followed the standard procedure of scale validation. First, a version of the scale with 10 items was translated into Portuguese and sent to three experts that are fluent in English and have knowledge about the subject. The experts were asked to make an analysis of the clarity, reading comprehension, general presentation and adequacy of the original meaning in English to Portuguese. The experts were instructed to first read the instrument in English. As one of the consulted experts was of Swiss origin, and therefore, also fluent in German, he added the German version in his analysis. All of them answered a specific evaluation sheet (analyzing each item), with the opportunity to make suggestions and improvements concerning the translation. There was a consensus among the specialists about adding some examples between brackets in order to clarify the meaning of "religious service" in the third item. The term is not culturally recognized as it is not part of Brazilians' vocabulary, and people would not understand its meaning. Depending on the individual's religious background a variety of terms like "worship", "mass", "sessions", "bible studies", "meetings", "prayer group", etc., could be employed to refer to what in English is stated as "religious service".
After incorporating the observations, a first Portuguese version of the scale with 10 items was created. Except for the first item (about the attendance to religious services, with seven seven-point answer scale) and the last one (about the frequency of prayer practice, with nine-point answer scale), the answers are given on a Likert scale of five points, measuring the level of interest/importance (1-not at all to 5-very much) and frequency (1-never to 5-frequently). The instrument was sent to the author of the scale, in a reverse translation from Portuguese to English. The author suggested the use of the original version in English (there is also an original version in German) and approved the reverse translation.
This first version of the scale was given to 42 college students who also served as a focus group. They discussed its semantic clarity and evaluated the words that could be understood by the general population, regardless of religion/belief. An equivalent version of the original scale was obtained, with appropriate adaptations to Brazilian culture. In the Brazilian version, the first and last items are closer to the German version, in which seven-point answer scale and nine-point answer scale are presented. For the analysis procedures, these alternatives were recoded to the Likert scale from 1 to 5 (according to Tables 2 and 3). After these procedures, the final version was applied to the general population. Table 2. Recoding of objective frequencies of the items concerning prayer, meditation, and religious services into five score levels. Adapted from Huber and Huber (2012, p. 720).

Coding of Frequencies of Religious Behaviors
Except for the items related to the attendance to religious services and frequency of prayer practice, the items on the scale are rated from 1 to 5 according to the Likert scale. For the calculation of the CRS score, the items are summed up and divided by the number of scores for each item. Thus, the classification results in the following parameters for the scale categorization: Highly Religious-4.0 to 5.0; Religious-2.1 to 3.9; Non-Religious-1.0 to 2.0.
Considering that the items on attendance to religious services and frequency of prayer practice are presented with a greater number of possible selected answers, these items were recoded from 1 to 5 (Table 2) for the calculation of the Likert scale. The alternatives presented in these two items consider the religious context of the Brazilian population. In this sense, the hermeneutic proposal of Huber and Huber (2012) is used for the recoding of these items (see Table 3).
Huber points out that the levels 1 and 2 about frequency ("rarely", "never") and importance ("not very much", "not at all") of religious services and prayer practice indicate that such religious constructs are almost absent from the individual. Consequently, they are not psychologically relevant, being categorized as Non-Religious. Conversely, individuals who indicate "very often" or "often" for the question on the attendance of religious services and "very much so" or "quite a bit" for the question on personal prayer practice, clearly express the presence of such constructs in their personal religious construct-system and evidence, therefore, that these elements have high psychological relevance for them, being plausibly categorized as Highly Religious. The "intermediary" level 3 presents responses, such as "occasionally" or "moderately", thus, indicating an infrequent and not very intense presence of religious elements that characterize the "religious" individuals.

Sample
All participants gave their informed consent for inclusion before participating in the study. The study was conducted in accordance with the declaration of Helsinki, and the protocol was approved by PUCPR Research Ethics Committee (Process 1.354.361, from 7 December 2015).
A link to an electronic inquiry developed in Qualtrics Online software was broadly distributed in social media inviting people to participate. A total of 796 individuals started the questionnaire, but only 723 completed it. Still, it was necessary to exclude 36 individuals as they were considered outliers (they completed the questionnaire in less than 1 minute), answering most of the questions in a pattern (same answers) or made inappropriate comments during the survey.
Therefore, the total valid sample of this study consisted of 687 participants from several different states in Brazil. Later, this sample was randomly divided into two different samples (approximately 50% each) for the Exploratory Factor Analysis (EFA-Sample 1) and the Confirmatory Factor Analysis (CFA) Sample 2). The majority of the total valid sample were women (64.9%) and had a high instruction degree (28.4% were college students, 26.8% had college degrees, and 35.4% had higher degrees-MBA, PHD, etc.), 7.6% had school degrees, and 1.2% did not complete school. Although this pattern indicates a sample of higher education above Brazilian standards, it can be explained by the form and means the data was collected with. Nevertheless, most of the sample had a general income below eight minimum wage. A minimum wage in Brazil is currently equivalent to the US $264.00. (19.2% below one minimum wage, 29.1% between one and three, 28.1% less than eight) and only 23.6% with income above it-similar to the Brazilian socioeconomic pyramid. Table 4 shows that, although women and men demonstrated no difference in their CRS score, some items reveal relevant differences between gender. In general, compared with men, women believe slightly more in God or something divine (item 02) having experiences where they felt God or something divine tried to communicate something (item 10), pray more (item 04) and think prayer is more important (item 09). Men, on the other hand, are more interested in learning about religious topics (item 06) and answered that they take part in religious activities more frequently (item 02). The random division of the sample resulted in a Exploratory Factor Analysis (EFA) sample (Sample 1) composed of 334 participants (48.62% of total valid sample), with a female majority (64.4%) and a Confirmatory Factor Analysis (CFA) sample (Sample 2) with of 353 individuals (51.38% of total valid sample), again with a female majority (65.4%).
There were no significant differences between samples regarding Age (M = 37.60 SD = 13.82) or regarding most of the questions from the CRS scale (all p-values non-significant, above 0.445). There was no statistically significant difference in the CRS score (the sum of all items divided by the number of items, Huber and Huber 2012, p. 720) between samples (p = 0.633) and its 10 items (the order and the translation of the items used in the questionnaire can be seen in Appendix A).
We did not find any significant difference regarding marital status (Married-36.0%; Singles-46.4%; As a marital status-9.6%, Separated/Divorced-6.3%; Widowers-1.7%) between Sample 1 and Sample 2. Both samples included participants from different religions, without significant differences between them (p = 0.719), with Catholics as the largest group (33.5%), Evangelicals as the second largest group (24.5%) and Spiritism as the third (11.9%). Although 15% of the sample said they believed in God, they said they had no religion. Furthermore, smaller percentages of other religions were reported: African-Brazilian, Pentecostal, Buddhism and combinations of different religions.

Correlations and Exploratory Analysis-Sample 1
Following Fradelos et al. (2018), for the conduction of the Exploratory Factor Analysis, a principal components analysis and Varimax Rotation method were both used. This choice was made due to the correlations found between items (shown below in Table 5) and in order to make comparisons between results.  Table 5 shows there are significant correlations between all items, indicating initially that the items would converge into a construct (the CRS).
The strongest correlations of each item (marked) are set between items formerly belonging to the dimension described by Huber and Huber (2012). For example, although item 05 ("How often do you experience situations in which you have the feeling that God or something divine intervenes in your life?") significantly correlates with item 02 ("To what extent do you believe that God or something divine exists?") or with item 09 ("How important is personal prayer for you?"), it has a stronger correlation with item 10 which belongs to the same factor ("religious experience") described by these authors.
The exploratory factor analysis (EFA) was conducted with the 10 items of the scale on a sample of 334 participants, that comply with the majority of the recommendations in the literature and also respects the minimum of 200 cases for any EFA (Guilford 1954;MacCallum et al. 1999;Hair et al. 2009). Results showed that the sample was adequate for a factor analysis (KMO = 0.904, Bartlett's sphericity x 2 = 2300.25 p < 0.001 df = 45). The first attempt used the K1 method proposed by Kaiser (1960), and eigenvalues analysis (Cattell 1966) presented a single factor solution with 60.27% of total variance explained. Still, it showed that a 05 factors-solution congruent with Huber and Huber (2012)-would explain 86.79% of the total variance. This was 05 factors solution with a Varimax Rotation and resulted in convergent results (Table 6) of the factors proposed by Huber and Huber (2012) with the exception of one item (02).
The Rotated Component Matrix table suppressing coefficients smaller than 0.4 presented that only item 02 from Ideology ("To what extent do you believe that God or something divine exists?") showed some different results from Huber and Huber (2012) as it should preferably be grouped with the Private Practice factor. We analyzed this factor grouping difference and discovered that it only happens on individuals who declared themselves as Catholics (about 1/3 of the sample), mainly because it revealed a strong correlation (r = 0.742 p < 0.001) with the item 09 ("How important is personal prayer for you?") (only a minor correlation with item 04, r = 0.386). Therefore, as item 02 was still grouped with the Ideology factor with a score higher than 0.40, we decided to maintain it in the Ideology dimension.
Due to these results, the CFA was tested with these concurrent models: (1) The CRS-10BR with five dimensions proposed by Huber and Huber; (2) the CRS-10BR with a single dimension, and; (3) the CRS-5BR within a single dimension.

Confirmatory Analysis
With results congruent with Huber and Huber (2012), we conducted a series of CFA analysis using the remaining half of the data (Sample 2). As with an EFA, a CFA can be used to reduce the overall number of observed variables into latent dimensions based on their commonalities. This is done by reducing measurement error and adding a level of statistical precision that is useful to scale validation (McArdle 1996).
The first model tested in AMOS v.20 for SPSS v.20 was the original CRS-10 with five dimensions (Intellect, Ideology, Public Practice, Private Practice and Religious Experience) model proposed by Huber and Huber (2012). It resulted in good fit indexes (χ 2 /df With this fit index, Hair et al. (2009) recommend that the model should be tested for its convergent and discriminant validity. Convergent validity tests if the items that form a dimension have a lot of common variances, and therefore, relates to the same dimension "as they converge to a common point". Discriminant validity assumes that one scale (or subscale) is different from other related scales (or subscales). Therefore, testing for discriminant validity between dimensions (and its inner items) is a way of testing if the dimensions are different from other dimensions of the same construct.
Convergent validity was established for all five dimensions based on AVE (Average Variance Extracted) values higher than 0.5, as recommended by Hair et al. (2009). Composite Reliability (CR) of all five dimensions were over 0.8, showing high reliabilities (Hair et al. 2009).
In the last step, the discriminant validity was confirmed as the AVE demonstrated in Table 7 is higher than the corresponding correlation coefficient between factors demonstrated in Table 8 (Fornell and Larcker 1981). Accordingly, it is possible to assume that, for this sample, the dimensions are or were perceived as different from the other dimensions of the CRS-10 scale. Also, general internal consistency for CRS-10BR was 0.921 (Sample 1 = 0.922/Total Sample = 0.922), again congruent with the authors, the internal consistency is highly acceptable.
As a final model, we tested the CRS-5BR with one dimension (Huber and Huber 2012). Similar to Huber and Huber (2012), our initial results for internal consistency using Cronbach's Alpha for CRS-5 were above acceptable (0.852) in the total sample (Sample 2: 0.862/total sample = 0.857), but not for all fit indexes (χ 2 /df = 5.636, NFI = 0.96, IFI = 0.97, TLI = 0.94, CFI = 0.97, GFI = not available, AGFI = not available, RMSEA = 0.115, SRMR = 0.03) (see Table A1 for the comparison of all models in the Appendix B). Therefore, it would be preferable to use the CRS-10BR with the five original dimensions rather than its CRS-5 counterpart.

CRS-Score Categorization Comparisons
We performed another comparison of the CRS-10 and CRS-5 by categorizing the individuals using Huber and Huber (2012) method. Individuals should be categorized as Highly Religious (CRS scores from 4 to 5), Religious (from 2.1 to 3.9) and Non-Religious (from 1 to 2) using the CRS score. First, we computed the average of 05 items (CRS-5-1 dimension) and last, with the average of the five dimensions (CRS-10-5 dimensions). Categorization results showed significant congruence with a general convergence of 90.5% of the cases (CRS-5: HR-65.6%; R-26.5%; NR-7.9%/CRS-10-HR-59.8%; R-32.2%; NR-8.0%/Kappa = 0.817, p < 0.001). Most of the differences (46 cases from the 65 not convergent) were due to individuals being categorized as "Religious" using CRS-10 instead of being categorized as Highly Religious using CRS-5 from the start. Paired t-test demonstrated similar results showing a high correlation between the CRS scores (r = 0.971, p < 0.001) and, although it presented different means-nonetheless it was a very small difference effect (M 5items = 4.01, SD = 1.02/M 5factors = 3.87, SD = 0.98, t(0.686) = −15.436, p < 0.001, d = 0.137).
Evangelicals, Pentecostals and Spiritists showed larger groups of Highly Religious individuals, followed closely by Catholics. Although 103 individuals declared they had no religion (but believe in God) more than 75% of them were categorized as Religious in both categorizations. Table 9 shows that there are few distortions between categorizations, mostly on Catholics and Spiritists. Still, our mean scores are a little lower than Huber and Huber (2012) (here M 5items = 4.01, SD = 1.02/M 5factors = 3.87, SD = 0.98-in their article M = 4.18, SD = 0.70) as our group categorizations were also a little different from surveys that previous presented in Brazil (p. 722) (NR = 2%/R = 24%/HR = 74%) by Huber and Huber (2012).
Crosstab analyses of the CRS categorizations between Educational levels demonstrated most of the same differences between Religious and Highly Religious (Table 10). The analyses revealed that most of the individuals were Highly Religious, followed Religious individuals. Nevertheless, people with Incomplete Graduation presented less Highly Religious and more Religious individuals than the other educational levels (CRS-5 p = 0.014/CRS-10 = 0.036) (13 individuals excluded because of fewer cases of other educational levels).

Conclusions
The validation process of the centrality of religiosity scale CRS-10BR and CRS-5BR shows adequate overall psychometric qualities for the use in the Brazilian cultural context. Exploratory and Confirmatory Factor Analyses indicated that the CRS-10BR shows better fit indexes than CRS-5BR; therefore, CRS-10BR is recommended. However, as the CRS-5BR has demonstrated acceptable fit indexes as well, it is also considered suitable for use in the Brazilian population when the context demands simpler and faster data collection, for instance, in healthcare settings. As noted by Hair et al. (2009), although a larger number of questions related to the same topic makes the scale more reliable, this would result in a longer scale hindering its applicability. Congruent to this, Huber and Huber (2012) point out that it is also important to look for the reliability of scales, specifically with fewer or one item measures (as "How important religion is for you?" or "How religious do you consider yourself?"). Indeed, this study demonstrated congruent results to Huber and Huber (2012, p. 716) that, the more items the CRS scale has, the more reliable and accurate it should be.
Nevertheless, both versions revealed a categorization convergence of 90.5% of the cases between Non-Religious, Religious and Highly Religious individuals. Thus, although it would be statistically recommended to use CRS-10BR, depending on the necessity to use shorter questionnaires and for later categorization using CRS-Score, it would be acceptable to use CRS-5BR.
The results indicated that the scales CRS-10BR and CRS-5BR are useful instruments for Brazilian reality, as it will allow researchers to make comparisons between data obtained in different scenarios, including research and the clinical context. Especially in the area of health, that has had a growth in research on spirituality/religiosity in Brazil; and it is useful to have a fast, short, and reliable instrument for data collection. In addition, identifying whether religiosity is central to the individual's life gives consistency to other data (for example, attachment to God, spiritual/religious coping, spiritual struggles, etc.), as pointed out by Huber and Huber (2012), "only if the religious construct-system is situated in a central position, religious beliefs can be powerful enough to influence subjective experience and behavior".
In the use of a self-applied instrument, there is the possibility of error in the self-understanding of the participant. In addition, responses may also be influenced by socially desirable patterns (Olson et al. 2016, p. 87). Therefore, the interpretation of the data should consider these aspects.
As a study limitation, the majority of the sample were women, but it is mostly due to the "prevalence of women in the Brazilian population, convenience sampling and the possibility of women being more prone to cooperate" (Esperandio et al. 2018, p. 6). Also, the majority of the sample had a high instruction degree, a pattern that is not congruent with Brazilian standards. Although it can be explained by the form and means the data was collected with, other studies can try to collect data with differently for example, in public spaces or even using other sampling techniques). For future research, it is necessary to understand the difference for dimension loading results on EFA for Catholic individuals, as for other categorization between religions and, if it could appear in other sociodemographic categories as Educational level, or even regional context form Brazil and Latin America. Moreover, with the validated scales (CRS-10BR and CRS-5BR) could be understood if, or even to what extent, it can be influencing individual lives in their behavior and decisions.