Confirmatory Factor Analysis of the Academic Self-Efficacy Scale: An Indonesian Version

Several studies regarding academic self-efficacy are developed in which a valid and reliable measurement is needed. One of the well-known instruments used to measure college students' academic self-efficacy is The Academic Self-Efficacy Scale (TASES). It was designed by Sagone and Caroli (2014), comprising four dimensions, i.e., self-engagement, self-oriented decision-making, others-oriented problem-solving, and interpersonal climate. This instrument contained 30 items at first, but two items were removed after testing the factor analysis, and 28 items remained. This study examined the validity of the adaptation of TASES into the Indonesian version. This scale was adapted into the Indonesian version using confirmatory factor analysis (CFA), involving 166 Indonesian college students studying at universities in Indonesia and abroad. The CFA results showed that the items which were distributed in 4 dimensions in this scale are found to fit except three items of interpersonal climate dimension. Therefore those three items have been eliminated. In addition, the coefficient of Cronbach's Alpha of TASES Indonesian version is highly reliable. Ultimately, the TASES Indonesian version consisting of 25-item within four dimensions has shown to be a reliable and valid measurement for academic self-efficacy in the Indonesian context.


Introduction
First-year undergraduate students often experience tension, at least moderate stress, because the academic pressure between university and secondary school is quite different (Pierceall & Keim, 2007). Freshmen students who have poor coping skills and are difficulty adjusting to the academic environment often encounter stress, low level of well-being, depression, anxiety, and even dropout (Garett et al., 2017;Sharma & Wavare, 2013;Owens, Stevenson, Hadwin, 2012;Stinebrickner & Stine, 2012).
A similar condition related to students' responses to the current academic situation was reported by Maulida, 2012. Maulida (2012 revealed that some students in Universitas Indonesia admitted to being depressed, characterized by cheerlessness or losing interest, worthless, frustrated, hopeless, and they also had suicidal thoughts. The data were obtained from a counseling service unit for college students in Universitas Indonesia (Badan Konseling Mahasiswa Universitas Indonesia) (Maulida, 2012). Furthermore, the Ministry of Research, Technology, and Higher Education (Kemenristek) of the Republic of Indonesia informed that in 2017 out of 6,924,511 enrolled students in state universities and private universities in Indonesia, 195,176 students (2.8%) dropped out (Talar & Gozaly, 2020). From those numbers, the percentage of dropout students in private universities was higher than in public universities, 4% and 0.3%, respectively (Talar & Gozaly, 2020).
The above cases result from a low sense of academic self-efficacy of the students (Preiss, Gayle & Allen, 2006;Richardson, Abraham & Bond, 2012;Robbins et al., 2004). Conversely, students with a higher level of academic self-efficacy believe that they are more capable of meeting challenges at school, tend to be more motivated, use more strategies (such as self-regulated learning), have greater achievement, and feel less tension and anxiety (Morton et al., 2014;Barrey & Finney, 2009). Therefore, they can easily adjust to the academic climate and socio-academic climate that influence their academic achievements.
The concept of academic self-efficacy referred to the concept of self-efficacy by psychologist Albert Bandura (1977Bandura ( , 1986. Albert Bandura (1977Bandura ( , 1986Bandura ( , 1997 has defined self-efficacy as one's belief in one's ability to succeed or accomplish a task in a particular situation and learn or perform behavior at designated levels. In the academic context, Schunk (1991) stated that self-efficacy is a person's belief that they can complete academic tasks successfully (e.g., taking notes, asking questions, etc.). Putwain et al. (2013) reported that academic self-efficacy could predict future academic performance in a university setting and predict learning-related emotions through academic performance. Li (2012) showed that the higher academic self-efficacy students have the more effort students put into the subject. Further, academic self-efficacy and academic achievement are also positively correlated (Li, 2012). Morton et al. (2014) demonstrated that a high level of self-efficacy will experience life with less stress in the first year in university.

120-132
This is an open access article under CC-BY-SA license (https://creativecommons.org/licenses/by-sa/4.0/) critical problems by using other people to support, and the last dimension (interpersonal climate) is the ability to create a prosocial and collaborative climate in interpersonal relationships (Sagone & Caroli, 2014).
CSEI (College Self-Efficacy Inventory) Solberg et al. (1993) constructed CSEI (College Self-Efficacy Inventory) to measure college students' selfefficacy associated with academic achievements. It was proposed to understand the role of self-efficacy of Hispanic college students in the process of college adjustment in the United States (Solberg et al., 1993).
CSEI consisted of 20 items with a 10-point Likert scale to rate the confidence. The higher point indicated the higher confidence to complete the task related to the college (Gore, 2006;Solberg et al., 1993). In addition, CSEI contained three categories: course self-efficacy (7 items), roommate self-efficacy (4 items), and social self-efficacy (9 items) (Gore, 2006;Solberg et al., 1993). Midgley et al. (2000) demonstrated that patterns of adaptive learning scales (PALS) were designed in educational settings to understand the correlation between learning environment and students' motivation, affect, and behavior. PALS for students consisted of 94 items within five broad components such as personal achievement goal orientation, perceptions of teacher's goal, perceptions of the goal structures in the classroom, achievement-related belief, attitude, and strategies, and perception of parents and home life . This scale used a 5-point Likert scale ranging from 1 (not at all true), 3 (somewhat true), and 5 (very true) .

PALS (Patterns of Adaptive Learning Scales)
While PALS for teachers consisted of 29 items, it evaluated teacher's perception of the goal structure in the school, goal-related approaches to instruction, personal teaching efficacy and also used a 5-point Likert scale ranging from 1 (strongly disagree), 3 (somewhat agree), and 5 (strongly agree) .

SEQ-C (Self-Efficacy Questionnaire for Children)
According to social-cognitive theory by Albert Bandura, Muris developed a self-efficacy questionnaire for children (SEQ-C) to measure the conviction of youth regarding their social, academic, and emotional domains (Muris, 2001). This scale was developed in the Netherlands, and the sample was limited to European youths aged 14-17 years old. It consisted of 24 items ranging from 1 (not at all) to 5 (very well). SEQ-C scores are associated with a measure of depression in a potentially significant manner. It can be seen that the lower the SEQ-C scores of the youths mean the greater level of their depression (Muris, 2001).

ABC (Academic Behavioural Confidence)
Sander and Sanders (2009) demonstrated that the Academic Behavioural Confidence (ABC) was designed to provide a global measure of academic confidence, and it can be beneficial for teachers in understanding their students and in knowing students' achievement levels (Sander & Sanders, 2009). ABC consisted of 24 items and each of which respondents rated on a 5-point Likert scale. This scale may arise from four basic concepts of self-efficacy: mastery experience, vicarious experience, verbal persuasion, and physiological states (Bandura 1977(Bandura , 1986(Bandura , 1993. No study has adapted the academic self-efficacy scale (TASES) developed by Sagone and Caroli (2014) into the Indonesian version. The academic self-efficacy scale (TASES) was used in the current study to investigate the perceived self-efficacy of first-year college students in an educational context in Indonesia. It is a critical step since an instrument's adaptation has a cultural fit to be used in various cultural contexts (Borsa et al., 2012). Furthermore, several studies have reported that language and cognition have a tight relationship. For example, King (2017) and Lupyan and Lewis (2019) reported that language impacts the cognitive structure and semantic knowledge. Furthermore, language can influence and shape one's thoughts, decision-making, and strategy for solving crucial problems (Gleitman & Papafragou, 2012;Novack & Goldin-Meadow, 2015).
Thus, adapting an existing instrument is needed to measure data from different samples, backgrounds, and even languages through several steps that provide greater fairness in the assessment (Borsa et al., 2012). Therefore, this present study attempted to validate the adaptation of TASES into the Indonesian version using confirmatory factor analysis (CFA) to assess the academic self-efficacy of Indonesian college students.

Methods
The adaptation process in this study referred to Beaton et al. (2000) that contained six essential stages: [1] Instrument translation from the original version into the Indonesian language; [2] Synthesis of the translated version from translators; [3] Back translation conducted by translators; [4] Expert committee (e.g., researchers, lecturers, or academic self-efficacy experts); [5] Test of the pre-final Indonesian version; and [6] Submission of documentation to the developers or coordinating committee for appraisal of the adaptation process.
Instrument translation into Indonesian version. Three translators conducted the translation process. The first and second translators are those who understand the concept of the academic self-efficacy scale and are English experts. The third translator is an English specialist but has less knowledge of the scale.

Synthesis of the translated version.
Researchers held a conversation with the translators after translating the instrument. Each of the issues in the translation results was discussed, and problems were resolved.
Back translation. After synthesizing the instrument, the scale was retranslated into the English version. In this process, two students in one of the colleges in Australia were involved. This validity testing method ensures that the translated version represents the exact item content as the original one. Furthermore, this stage is fruitful in controlling the validity and identifying major inconsistencies and conceptual errors in the translation (Beaton et al., 2000).
Expert committee (researchers, lecturers, or academic self-efficacy experts). Peer judgment and expert judgment were needed in this step. In this process, peer judgment includes two people, both researchers, and lecturers in the Faculty of Psychology at one of the universities in Indonesia. Additionally, the expert judgment involves a researcher and a lecturer in the Faculty of Psychology who has a better understanding of the concept of academic self-efficacy and has been publishing academic papers about academic selfefficacy.
Peer judgment and expert judgment have reviewed the scale's translation, synthesis, and backtranslation. Some critical points need to be noted, which are: semantic equivalence, after translation, synthesis, and back-translation have been completed, each item should have the same semantic meaning as the original version. Idiomatic equivalence, idioms are typically difficult to translate since they are based on the cultural context of the original developers. Consequently, the appropriate expression in the Indonesian version should be formulated by the committee. Experiential equivalence, each country has different daily life experiences, so each item of the instrument should be replaced by a comparable item that is currently practiced in Indonesian culture. Conceptual Equivalence, distinct conceptual meanings are often found among Test of the pre-final Indonesian version. This part consisted of the psychometric process. We split this part into some points:

Procedure.
After considering the modifications suggested by the experts and receiving informed consent letters from the participants, researchers assessed the Indonesian version of the academic self-efficacy scale (TASES). The participants were collected through convenience sampling (Cozby & Bates, 2015;Gravetter & Forzano, 2018). Google Form was distributed to the target participants through social media such as WhatsApp, Facebook, and Instagram. Sagone and Caroli (2014) to measure the academic selfefficacy of college students based on the self-efficacy theory of Albert Bandura. It consisted of 30-item within four dimensions. After validating the instrument using an analysis factor approach, 28-item left within four dimensions. Those four dimensions are self-engagement, self-oriented decision making, others-oriented problem solving, and interpersonal climate (Sagone & Caroli, 2014).
TASES is a self-report scale using a 4-point Likert scale. The original Likert-type scale of this instrument was 7-point. Yet, it changed into a 4-point Likert scale after the peer judgment process to minimize the participants' neutral response option.
Construct validity has been used in this present study to examine the validity (Cohen et al., 2013) with analysis factor (confirmatory factor analysis/CFA; Cohen et al., 2013). To analyze with CFA, the researchers used structural equation modeling (SEM) with R application 4.0.2 version (2020-06-22) in which Lavaan 0.6-6 version has been installed. In order to answer the research questions of this study, TASES is indicated fit when the values meet the criteria as follow: CFI > .95, RMSEA < .06, SRMR < .08, TLI ≥ .95 (Hu & Bentler, 1999). However, the Chi-Square index (c 2 ) has not been reported due to its sensitivity to the sample size (Lacobucci, 2010;Brown, 2015;Little, 2013). On the other hand, the items should be eliminated based on the factor loading value (l) £ .32 (Tabachnick & Fidell, 2007). Lastly, AIC (Akaike Information Criterion) is also reported indicating that smaller values are better for comparing the model (Yamin & Kurniawan, 2009;Akaike, 1974).
Reliability test was performed by comparing the coefficient of Cronbach alpha (Cohen et al., 2013) with the coefficient alpha ranging from .70 to .90 and indicating a high-reliability score (Hinton et al., 2004). In other words, the coefficient alpha value is above .70 (Bland & Altman, 1997). In this present study, IBM SPSS 24 version has been used to test the reliability.
Submission of documentation to the developers or coordinating committee for appraisal of the adaptation process. This process was supported by the developer, who is currently a lecturer and researcher in the Faculty of Psychology in Indonesia, and attended by several graduate students of the Faculty of Psychology in Indonesia. They conducted a review that started from translating to validating the instrument. Furthermore, they also recommend some suggestions and comments regarding the adaptation process based on the Indonesian version.

Results and Discussion
The mean and standard deviation of the scores of the four dimensions of the TASES Indonesian version are presented in Tables 1 and 3. The score mean and standard deviation after the validation process are reported in Table 1, while Table 2 provides the scores before the TASES Indonesian version validation. As reflected in Table 1 and Table 2, mean and standard deviation scores showed the same results except for the interpersonal climate dimension. Thus, after validation, three items were eliminated from that dimension, and the scores of mean and standard deviation became higher. Furthermore, Table 3 presented the factor loading of each item and Cronbach's α as a whole and per dimension after validation (containing 25 items). Meanwhile, Table 4 showed the factor loading of each item and Cronbach's α for the whole scale and per dimension before validation, containing 28 items as the original version.
Before the TASES Indonesian version involving 166 Indonesian college students (studying at universities in Indonesia and overseas) was validated, it showed poor fit results. The items that have been allocated in four dimensions (i.e., self-engagement, self-oriented decision-making, other-oriented problem-solving, and interpersonal climate) declared fit with indices mentioned above except for three items of the interpersonal climate dimension. Therefore, those items have been eliminated.
Subsequently, after TASES Indonesian version was validated, its results showed a good fit with the scores of RMSEA = .078, CFI = .784, TLI = .761. Referring to the standard scores of RMSEA, SRMR, CFI, and TLI mentioned by Hu and Bentler (1999), the measurement is declared to be valid based upon some criteria, CFI > .95, RMSEA < .06, SRMR < .08 (Hu & Bentler, 1999). Thus, the first results of this measurement indicated poor fit. Hu and Bentler (1999) stated that the objective of the cutoff criteria, RMSEA, TLI, and Mc, is to reject populations with a small number of participants. A small population means the sample size is below 250 (Hu & Bentler, 1999). As in this study, the sample size is < 250; thus, it is not recommended to use the cutoff criteria as a reference for the fit index.
The IC score after the validation process was smaller than the score before validation. This smaller score proves that the parsimony is better for comparing the two models (Akaike, 1974;Yamin & Kurniawan, 2009). In short, the second model was found to be fit (i.e., parsimonious/adjusted fit measures). Furthermore, Yamin and Kurniawan (2009) stated that RMSEA < .08 is a good fit (i.e., absolute fit measures), and in this current study, we found the RMSEA score = .078. Hence, referring to Yamin and Kurniawan (2009), the Indonesian version of TASES fits with RMSEA < .08.
Meanwhile, the incremental/relative fit measures used the TLI (i.e., Tucker-Lewis Index) and CFI (i.e., Comparative Fit Index) score criteria (Yamin & Kurniawan, 2009). In Yamin and Kurniawan (2009), the provisions of the TLI and CFI scores showed that .80 ≤ TLI ≤ .90 and .80 ≤ CFI ≤ .90 are marginal fit. In this study, we found that TLI and CFI scores were .761 and .784, respectively. If the two scores are rounded off, the acquisition of both TLI and CFI scores is .80. Thus, on the incremental/relative fit measures, the Indonesian version of TASES was found to be a marginal fit.
Internal consistency is used to assess that the instrument has a reliable measurement function (Cohen et al., 2013). We applied the coefficient alpha (Anastasi & Urbina, 2016) developed by Cronbach (1951) to test the internal consistency. The Cronbach's αcoefficient of TASES Indonesian version was found higher (α = .893) than the original version (α = .880; Sagone & Caroli, 2014). This coefficient indicated that TASES Indonesian version is proven to have a high consistency level (α < .70) in measuring academic self-efficacy in different participants (in the original version, the participants were Italian college students).
In this case, self-engagement drives students to concentrate on lectures by participating in lecture activities (Sagone & Caroli, 2014). This is how they can observe the lecture process, understand the course topic, and ask questions (for instance) to improve their lecture's understanding.

127-132
http://journal.uinjkt.ac.id/index.php/jp3i This is an open access article under CC-BY-SA license (https://creativecommons.org/licenses/by-sa/4.0/) Self-oriented decision-making refers to how individuals only depend on themselves in the face of an unpleasant lecture process (Sagone & Caroli, 2014). Even though students choose a study program according to their interests, there are still some cases of students selecting a study program not following their passion and interest (Prabowo et al., 2019). Each of them has their challenges during the lecture periods. These challenges include the ability to manage time, complete lecture assignments, be active in organizational activities, and so on.
It is certainly not easy for students who have to complete multiple college tasks at nearly the same time. But, on the other hand, they are required to accomplish their courses that they are not interested in.
In this case, they should have strategies to organize themselves and to overcome the difficulties during the lecture process.
Others-oriented problem-solving is related to the role of the involvement of others (such as friends, lecturers, academic staff, and so on) in solving problems/obstacles experienced related to lectures (Sagone & Caroli, 2014). In this case, students cannot be separated from several things that hinder the lecture process. Thus, others-oriented problem-solving is seen when students maintain good relationships with friends and establish good communication with lecturers. When they notice something they do not understand, they will ask for an explanation from the lecturer instead of looking for answers by reading the literature by themselves.
The last dimension is the interpersonal climate that focuses on how individuals can cooperate with their friends (Sagone & Caroli, 2014). The situation where students collaborate in group activities/assignments will make it easier for them to undergo the lecture process, especially in completing lecture assignments.
Self-efficacy, grounded from social cognitive theory by Albert Bandura (Bandura, 1986), is linked to many particular domains such as academic, social, career, clinical, and health areas (Bandura, 1997). In the educational context, when one has higher levels of academic self-efficacy, one tends to be more inspired, motivated, use more strategies (such as self-regulated learning) to achieve more achievements, and experience less tension and anxiety (Barry & Finney, 2009). In particular, self-efficacy has been extensively studied in academic and social fields with college-aged populations since they are the critical elements of academic life. Therefore, Sagone and Caroli (2014) constructed The Academic Self-Efficacy Scale (TASES) to measure academic self-efficacy.
Future research may consider creating new interpersonal climate dimension items of the TASES Indonesian version and should subsequently analyze them since three items have already been deleted. Therefore, TASES Indonesian version might be a more acceptable fit to assess the academic self-efficacy of Indonesian college students. Furthermore, the following researchers also may conduct comparative research regarding the differences between the academic self-efficacy of rural college students and that of urban college students and how technology can influence their academic self-efficacy. More additional research is needed.

Conclusion
The original version of the academic self-efficacy scale (TASES) consisted of 30 items at first within four dimensions. Then two items were removed after testing the analysis process, and 28 items remained. This present study attempted to adapt TASES into the Indonesian version using CFA and found that 25 items fit the criteria indices. In contrast, items 12, 24, and 25 did not fit the criteria and were eliminated. Thus, the Indonesian version of TASES also indicates high reliability and fits within four dimensions. In sum, the current version of TASES is a reliable and valid measurement instrument for academic selfefficacy for the Indonesian population. This research process used the Indonesian version of TASES by involving Indonesian students as participants. This indicates that, for Indonesian undergraduate and postgraduate student participants, the 25 items from TASES are useful for accessing students' confidence in their academic abilities during pursuing their degree.