The sample comprised 1,132 individuals (54.90% male and 45.10% female) from various regions in Spain. These individuals were recruited using intentional sampling. The sample had an age range of 18 to 89 years (M = 42.43, SD = 18.69). Overall, 663 individuals were from urban regions and 469 were from rural regions (with rural regions defined as those with populations of 2,500 inhabitants or fewer, and urban zones with populations greater than 2,500 inhabitants). To be eligible for the study participants had to be adults, Spanish by nationality and could not have any form of physical or mental disability.

The Healthy Lifestyle Questionnaire (Cuestionario de Estilos de Vida Saludables; CEVS-II). The 27 items of this instrument are grouped into seven factors: Balanced diet (3 items; e.g., "I normally eat fish 2 or more times a week"), Respect for meal schedules (3 items; e.g., "I usually respect the schedule of the main meals of the day"); Rest habits (3 items; e.g., "I usually sleep 7 or 8 hours a day"); Tobacco consumption (3 items; e.g., "I feel good when I smoke"); Alcohol consumption (5 items; e.g., "I think alcohol makes me feel good"); Consumption of other drugs (5 items; e.g., "I find it easy to get some drugs") and Physical activity (5 items; e.g., "I consider myself a physically active person").

To measure the level of self-determined motivation we used the Behavioral Regulation in Exercise Questionnaire (BREQ-3); Wilson et al., 2006), which was validated for the Spanish context by González-Cutre et al. (2010). This questionnaire consists of an introductory phrase "I exercise physically…", followed by 23 items that capture different types of motivation: Intrinsic Regulation (4 items; e.g., "…because I think physical exercise or sport is fun"), Integrated regulation (4 items; e.g., "…because I consider physical exercise or sport to be part of me"), Identified regulation (3 items; e.g., "…because I value the benefits of physical exercise or sport"), Introjected regulation (4 items; e.g. "…because I feel guilty when I don't practice it"), External regulation (4 items; e.g. "…because others tell me I should do it") and Amotivation (4 items; e.g., "…I don't see why I have to). This factor structure was shown to have adequate fit: SRMR = .06; CFI = .91; RMSEA = .06; χ2/df = 3.20.

To measure intention to be physically active we used the Measure of the Intentionality of Being Physically Active (MIBPA), which was developed by Hein et al. (2004) and validated in Spanish by Moreno-Murcia et al. (2007). This questionnaire is formed of five items that load on a single factor: intention to be physically active (e.g., “I am interested in developing my physical form”). This factor structure was shown to have good fit: SRMR = .04; CFI = .97; RMSEA = .08; χ2/df = 1.80.

Responses to the questionnaires were recorded on a five-point Likert-type scale that ranged from 1 (Totally disagree) to 5 (Totally agree) with the formulation of the item.

The study was granted approval by the Commission of Bioethics and Biosafety of the University of Extremadura (Spain), following the recommendations from the Helsinki declaration. All participants were treated in accordance with the American Psychological Association's ethical guidelines regarding participant consent, confidentiality, and anonymity.

Once the objectives of the study and measures were defined; the main instrument, the CEVS-II, was developed following the recommendations of Muñiz and Fonseca-Pedrero (2019). This instrument comprised the 12 items for the factors of the Leyton et al. (2018) questionnaire and additional items to measure the dimensions alcohol consumption, consumption of other drugs, and practice of physical exercise. For each of these new factors we proposed an initial six items, which were then analyzed by a group of five experts in Physical Activity and Sport Sciences, and two Psychologists, resulting in five items per new factor. Ultimately, the CEVS-II questionnaire comprised a total of 27 items.

Finally, a battery of the different questionnaires was prepared, including additional measures of sociodemographic variables such as gender, age, and region. The battery of questionnaires was then uploaded to the Google Form platform. Participants completed the battery of questionnaires individually online, accessed through different mediums (WhatsApp, Facebook, Twitter, and email). It took approximately 10-15 minutes to complete all questionnaires. The online form also included information about the investigation, informed consent, and instructions on how to respond to the questionnaire. It also included a contact phone number and email address to resolve any questions. No doubts or missing values were reported.

First, we used confirmatory factor analysis (CFA) to test structural validity. Because the endogenous variables were ordinal, we used a robust weighted least squares estimators (WLSMV), which is designed specifically for ordinal data (Li, 2016). For each latent factor, the loading of the first element was restricted to 1. We report standardized estimates despite using unstandardized parameters in the model. To evaluate the goodness of fit of this model we used (a) the χ2/df ratio, which is recommended to be ≤ 5 (Schumaker & Lomax, 2010); the comparative fit index (CFI), which is recommended to be ≥ .95 (Cangur & Ercan, 2015); (c) the root mean square error of approximation (RMSEA), for which values < .08 indicate acceptable fit (Browne & Cudeck, 1992); and (d) the standardized root mean square residual (SRMR), for which values < .08 indicate acceptable fit and < .05 indicate good fit (Schermelleh-Engel et al., 2003). We used a multigroup CFA approach to test the psychometric equivalence of EVS across (a) gender, and (b) age group. For age group we split the sample into two groups representing older (> 40 years) and younger (≤ 40 years) participants using a median split method. For this analysis we used the MLR estimator. In this approach, testing nested models with increasing numbers of parameter restrictions allows for an assessment of configural, metric, and scalar invariance. Configural invariance tests the extent to which the model fits across the global sample. Metric invariance (with loadings constrained across groups) tests the extent to which items load on the factors similarly across groups. Finally, scalar invariance (with loadings and intercepts constrained across groups) tests whether sample means can be meaningfully compared across groups. We determined invariance by examining changes in various alternative fit indices because these are typically less sensitive to sample size (Putnick & Bornstein, 2016). Typically, a criterion of a -.01 change in CFI paired with change in RMSEA of .015 and SRMR of .030 for metric invariance or .015 for scalar invariance is used as an indication of invariance (Chen, 2007).

We performed a descriptive analysis of scale means, skew, and kurtosis, as well as an analysis of internal consistency via the calculation of omega coefficients (McDonald, 1999). Finally, to determine concurrent validity, we calculated bivariate Pearson’s correlations between the factors of the CEVS-II and those of the BREQ-3 and MIBPA.

We conducted the CFA analyses using the lavaan package (Rosseel, 2012) in R. MPlus 7.11 (Muthén & Muthén, 2014) was used to test the seven-factor model. We used SPSS 21.0 (IBM-SPSS Inc, Chicago, IL) to perform descriptive analysis and calculate correlations.

According to fit indices, the correlated six-factor model had acceptable fit: SRMR = .059; CFI = .973; RMSEA = .049; 90% CI [.046 - .052]; χ2/df = 3.76. As can be seen in Fig. 1, the standardized factor loadings had moderate to strong values, ranging from .57 to .97. Most of the latent factor correlations were weak, with the exceptions of Balanced diet with Respect for meal schedules (.65), Balanced diet and Rest habits (.56), Respect for meal schedules and Rest habits (.68), and Alcohol consumption and Consumption of other drugs (.61).

Figure 1.

Confirmatory Factor Analysis. Circles represent latent variables (factors) and rectangles represent indicators (items).

(0.51MB).

Values for the fit indices in each of the multigroup models are shown in Table 1. Overall, the changes observed for CFI, TLI, RMSEA, and SRMR across models supported scalar invariance in males and females. Scalar measurement invariance in older and younger participants was only supported partially after freeing the constraints on intercepts for items 3, 17, and 22.

According to the rules of normality proposed by Curran et al. (1996), all the variables had univariate normality, as shown by skew values < 2 and kurtosis values < 7. McDonald’s omega coefficients are shown in Table 2.

The output of the correlational analysis is shown in Table 3.