Academic motivation scale - reliability and validity evidence among undergraduate nursing students*

Objective: to assess the evidence of validity and reliability of the academic motivation scale (AMS) based on the internal structure. Method: this is a methodological study with 205 undergraduate nursing students. Dimensionality/internal structure of the AMS was assessed using factor analysis in the context of exploratory structural equation modeling (ESEM) and reliability of the factors was assessed by Cronbach’s alpha (α) coefficient and composite reliability (CR) coefficient. Results: acceptable fit indexes were obtained (CFI = 0.92; RMSEA = 0.07; SRMR = 0.06) for a three-dimensional model: intrinsic motivation (10 items; α = 0.84; CR = 0.86); extrinsic motivation (8 items; α = 0.84; CR = 0.90); and demotivation (4 items; α = 0.84; CR = 0.88). A significant correlational pattern was found for the motivation continuum. Conclusion: the dimensionality analysis for the AMS presented a model with three factors: intrinsic motivation, extrinsic motivation and demotivation, and was considered a reduced alternative to the original version of seven factors. This study helped assess the validity of the measurement instrument and its theory refinement; further studies should be conducted to assess its invariance property.


Introduction
According to the self-determination theory (SDT), motivation can be analyzed as a complex and multidimensional theoretical construct. The types of motivation can explain the different reasons why people act in a certain way, based on different types of regulation and causal locus of behavior. Motivation is considered autonomous when an action takes place due to a genuine interest in the activity, while the most controlled types of motivation occur when there is internal or external pressure to engage in an activity. The SDT also includes demotivation, that is, the absence of any type of motivation (1)(2)(3)(4)(5) .
Based on the SDT, Canadian researchers developed the Echelle de Motivation en Education (EME) in French (6) , which was translated into English, validated and renamed as the Academic Motivation Scale (AMS) (7) . This scale was developed to measure the motivation level of students and their self-perception of the reasons for engaging in an activity (6)(7)(8) . In the academic context, students are expected to have the most autonomous types of motivation, as studies have demonstrated positive relations between these autonomous types and student performance (3)(4)(5) .
In comparison with the theoretical construct of the SDT, the scale included seven factors of specific taxonomy for the educational context: intrinsic motivation to know, to accomplish things, and to experience stimulation; extrinsic motivation by external, introjected, and identified regulation. Factor seven refers to demotivation. Extrinsic motivation by integrated regulation was suppressed in the AMS because, in the factor analysis, it is not distinguished from motivation by identified regulation (7)(8) .
In this sense, the first assessment of the AMS dimensionality showed inconsistencies in the exploratory factor analysis (EFA) and did not reproduce the theoretical model with seven subscales, with eigenvalues above one in at least one of the factors (6) . In the scale applied in Canada, the CFA admitted the 7-factor structure of the theoretical model inadequately reproduced the observed covariance matrix; however, fit indexes increased after the inclusion of 26 residual correlations based on the modification indexes (7) .
Later, the AMS showed concurrent validity with other motivation measurement instruments, with an appropriate correlation between the subscales regarding the hypothesis of motivation continuum postulated in the SDT, except for intrinsic motivation for stimulating activities, which showed a weak correlation (8) .
In Brazil, the scale was translated into Portuguese (23) and applied to medical students; however, no robust analysis has assessed its dimensionality and reliability as an instrument to measure motivation. Although accepted in national studies (15,17,21) , since it was developed, new guidelines for the assessment of psychometric properties, translation and adaptation of instruments have been incorporated into the literature (24)(25) . Therefore, processes must be adopted to evaluate its properties to ensure a valid and reliable scale for the measurement of academic motivation.
The stages of validation and evaluation of psychometric instruments must be performed in a continuous process to provide valid current reliable data about the measurement instruments (24) . In view of the above, this study aimed to assess the evidence of validity of the academic motivation scale based on the internal structure (26) and the reliability indexes of the proposed dimensions of measurement.

Method
This is a methodological study conducted with undergraduate nursing students from a public university in the state of São Paulo in the second half of 2014. It used a convenience sample, which consisted of students who were available to participate in the investigation.
Considering the scarcity of evaluations about the factor structure of the AMS in the Brazilian context, the author decided to use the version translated from English (7) into Portuguese (23) , maintaining all items of the original version. The authors of this instrument authorized to use it for educational purposes (7) .
The original version of the AMS proposes to analyze motivation in the academic context, covering 28 propositions divided into seven subscales with four items each, scored on a Likert scale from 1 (no agreement) to 7 (total agreement), with a mean of 4 (moderate correspondence) (7,23) . A student indicates the agreement with a statement, allowing the calculation of each type of motivation based on the items linked with the proposed theoretical construct, according to Table 1   Source: Developed by the author based on the 7-factor model originally proposed by Vallerand, et al. (6,7) and translated into Portuguese by Sobral (23) In addition to the score for each of the seven types of motivation, three different factors were defined at a higher order of the scale (6)(7)(8)17,21) : Intrinsic Motivation, Extrinsic Motivation, and Demotivation, based on the arithmetic mean calculation.
Data were collected at formal times of academic activity after obtaining the authorization of coordinators and professors of the course who indicated the best moment for the researcher to explain the study participation. The students who agreed to participate remained in the room to answer the instrument and sign an informed consent form (ICF), which was returned to the researcher. The instrument had no student identification to ensure data confidentiality.
Data were organized in Microsoft Office Excel spreadsheets and factor analyses were performed using the exploratory structural equation modeling (ESEM) (27) , whose source of information was a polychoric correlation matrix. The methods of weighted least squares means and variance (WLSMV) estimation and GEOMIN oblique rotation were adopted. These analyses were conducted using the Mplus 7 (28) and the number of factors to be extracted was indicated by parallel analyses (29) . The criterion for maintaining an item in the instrument was defined a priori: saturation of factor load ≥0.40 and item-total correlation ≥0.40.
Estimated factor solutions were evaluated based on theoretical reasonability, interpretation of factors according to theoretical assumptions (7) , and degree of factor model adjustment to the empirical data. The following criteria were considered: a comparative fit index (CFI) above 0.90 indicated good adjustment; root-mean-square error of approximation (RMSEA) and standardized root-mean-square residual (SRMR) values below 0.08 indicated adjustments in these two residual indices, and values below 0.06 (30)(31)

Results
This study analyzed 205 instruments, which were After the third factor, any values in a random data matrix would be able to produce eigenvalues above the empirical eigenvalues.
Considering this indication, factor analyses were performed with extraction of one, two and three factors.
The adjustment results of these models were: one-   Therefore, regarding the internal structure, the AMS could be represented in this study by three factors. Table   2 shows the estimated psychometric parameters for the items/factors.
When analyzing the items grouped by AMS factors based on the original proposal presented in Table 1, each factor corresponded to a different type of motivation:  The result of parallel analyses did not support the extraction into seven factors according to the theoretical proposal of the authors of the scale (6)(7) , since the matrix of empirical data of this study could be reduced to a maximum of three factors based on criteria found in the literature (29,35) .
The factor analyses in this study partially reproduced the theoretical model. Most of the item/factor relationships foreseen in the higher order structure (6)(7)(8) were observed and the factor loads estimated at the level of p ≤0.05 were considered "good" (above 0.55) and "excellent" (above 0.71), according to the proposed taxonomy (36) .
As seen in the results of this study, other studies also proposed a new reconfiguration of the AMS by EFA, with the extraction of four factors: demotivation, extrinsic motivation by external regulation, extrinsic motivation by identified regulation, and intrinsic motivation, confirming the results by CFA, with proper validity and reliability and significant losses in models of one, two and three factors (39) . Other studies also pointed out inconsistencies in the number of factors, with extraction of five factors, in which the three types of intrinsic motivation were grouped into a single factor (20,40) , and extraction of three factors (41) .
In a study with Lebanese medical students, PCA was used and the results showed the items converged into a 3-factor solution with explanation of 81.51% of total variance (11) .
In an assessment of the psychometric properties of the instrument with Chilean students of Dentistry (10) and students of the health vocational course and social care course in Norway (9) by CFA, the researchers confirmed a 7-factor model. In Argentina, two versions of the instrument were applied with high school and university students, and the results also reproduced the 7-factor model (14) . In a revised version in Spanish (16) that included the subscale extrinsic motivation by integrated regulation in the AMS from the original version (6) , applied to Pedagogy students by CFA, the results showed acceptable adjustment indexes of the new structure, with eight factors (16) . The authors proposed the inclusion of a factor previously suppressed in the AMS (7,8) : extrinsic motivation by integrated regulation, to obtain an instrument that would allow the measurement of all motivation regulations proposed by SDT (1)(2) in the educational context in Spain.
Regarding the pattern of item grouping, foreseen in the theoretical model in three factors of a higher order,  (42) .
Regarding the existence of the self-determination continuum and motivation (1,(6)(7)(8) seen as a significant positive correlation between intrinsic and extrinsic motivation and a negative correlation between intrinsic motivation and demotivation, the results confirmed the findings of other studies (15,17,22) and the theoretical model (1,2) . However, in the English version of the scale, most studies did not confirm this hypothesis (6)(7)(8)(18)(19)39) , suggesting this construct pattern did not exist or was limited (19,39) .
This study also assessed to what extent each modeled latent variable (that is, the AMS factors) was different and how it could be distinguished from the others. One way to obtain this evidence is to assess whether the square root of VME (√VME) is greater than the estimated variance shared between the constructs, i.e., greater than the correlation between latent factors.
The results in Table 3 show evidence that all three factors were properly discriminated in this study, according to the literature (43)(44) .
Regarding the content of the items, although it was not the objective of this study, the author of this study agreed with other authors who suggested that such content should be reviewed according to criteria recommended in the literature in order to make it more precise (17) and that the scale has to be revised and updated as a measurement tool (15,(18)(19)22) . Although it is still one of the main instruments of academic motivation measurement today (45,46) , this study found the scale has to be improved, since the measurement attributes did not fully correspond to the theoretical attributes.
Specifically in the Portuguese version for the Brazilian context, the AMS should be explored more rigorously in the process of translation, cross-cultural adaptation, equivalence and frequency of terms, including reverse translation, not previously described (23) . A simple translation of the items does not ensure the maintenance of the original construct. Future studies could include a sequence with a cross-cultural adaptation and a more robust analysis of the factor structure (25) .
In the other countries where translated versions were used or in the contexts where the English version of the AMS was applied, the adoption of other analysis resources is suggested, that is, besides the CFA, ESEM is recommended as it allows the integration of structural equation modeling and EFA in a single analysis (27) .
Through ESEM this study presented different outcomes from previous studies that assessed the psychometric properties of the AMS with the determination of a 3-factor model according to the reference in the subject (27) . diversified samples are required so that the invariance properties of the scale can be properly investigated.