Results and discussion PROS Nadi S, Te Sheng C Research on University fulltext

Research on University Student’s self-efficacy scale in science education: A systematic review SWUP SC.122 Research in Science Education--Res Sci Educ, and International Journal of Innovation in Science and Mathematics Education IJISME. Res Sci Educ was published by Springer and IJSE was published by Routledge Taylor and Francis Group. IJISME was published by University of Sidney, Australia. In addition, two of these journals listed in Social Science Citation Index by Thomson Reuters. In the following step, about eight articles within the last 10 years remained worthy of being read more closely. From the obtained collection, those articles were selected that present empirical studies, including self-efficacy scale, validity, and reliability in the sense identified above. Some authors stated clearly the set of their fields; some did not provide any information at all. Finally, the papers were chosen as listed in Table 1. Table 1 . Papers were selected in this study. No Author Title Name of the Journal Publisher 1 Uzuntiryaki Aydın 2009 Development and Validation of Chemistry Self-Efficacy Scale for College Students Research in Science Education Springer 2 Lin, Liang, Tsai 2014 Identifying Taiwanese University Students’ Physics Learning Profiles and Their Role in Physics Learning Self-Efficacy Research in Science Education Springer 3 Lin, Liang Tsai 2014 Conceptions of Memorizing and Understanding in Learning, and Self-Efficacy Held by University Biology Majors International Journal of Science Education Routledge 4 Lindstrøm Sharma 2011 Self-Efficacy of First Year University Physics Students: Do Gender and Prior Formal Instruction in Physics Matter? International Journal of Innovation in Science and Mathematics Education University of Sidney

3. Results and discussion

3.1 Results and discussion Previous researchers developed the self-efficacy item scale in science major and how they analyzed their data. They used a Likert scale questionnaire in their instrument. Most researchers reported that rating scale as interval data in published research Privitera, 2014. This type of scale is a numeric response scale used to indicate a participant’s level of agreement or opinion with some statement. Therefore, Self-efficacy scale are discussed here as an interval scale. This situation in-lined with some scholars who noticed that for using factor analysis of the item at least the variables should be the ordinal level of measurement Brace, et al., 2006. Granting to the selection of study method above, four papers from science education view were deduced from the empirical data. Each of the paper focuses on physics, chemistry, and biology; two papers in the physics domain, while others come from the fields of biology and chemistry. In addition, two papers were conducted in university in Taiwan while the others in Turkey and Australia. Based on Table 2, the previous researchers develop the self-efficacy item scale in science major and how they analyze their data. Looking at the number of participants, all researchers used a large sample more than 250. It is intended to obtain the stability of the data. Particularly, Lindstrøm Sharma 2011 used participant between 122 and 281 each time over a year due to their used Cohort study in their methodology. As a general rule, most item parameters can be estimated with relative N. Suprapto, T.-S. Chang SWUP SC.123 stability for samples of 200 participants, and so this might consider the minimum number desired Crocker Algina, 1986. Table 2. Overview about articles analyzed in this review. Study and Country Major Level and Total number of participants Instrument: Questionnaire Name of Instrument Dimension Number of items Uzuntiryaki Aydin 2008 Turkey Chemis College first phase= 363 students Second phase =353 students Chemistry Self-Efficacy Scale CCSS a self-efficacy for knowledge comprehension-level skills, b self-efficacy for higher-order skills, c self-efficacy for psychomotor skills, d self-efficacy for everyday applications Original 22 items, Finally 21 items after content validation Lin, et al 2014a Taiwan Physics Undergraduate2 50 students The physics learning self- efficacy PLSE a Conceptual Understanding 5 items b Higher-order cognitive skills 6 items c Practical work 7 items d Everyday application 8 items e Science communication 6 items 32 items Lin, et al 2014b Taiwan Biology 18 -master students; 82 undergraduate level; total 293 students Students’ biology learning self- efficacy BLSE a Higher-order cognitive skills, BLSE- HC 6 items b Everyday application, BLSE, EA 8 items c Science communication, BLSE-SC 6 items d Practical work, BLSE- PW 7 items 27 items Lindstrøm Sharma 2011 Australia Physics Undergraduate N between 122 and 2814 times in one year to the 1st year physics cohort The Physical Self-Efficacy Questionnaire Only 1 factor dimension since the author only focused on Physics self-efficacy items 5 items Research on University Student’s self-efficacy scale in science education: A systematic review SWUP SC.124 Table 3. Summary of data analysis Uzuntiryaki Aydin, 2008. Kind of analysis and the results Dimension Content validation Factor Analysis and T he total percentage of variance extracted Internal consistency from Cronbach’s alpha a self-efficacy for knowledge comprehensio n-level skills SCS b self-efficacy for higher- order skills an added dimension after EFA c self-efficacy for psychomotor skills, SPS d self-efficacy for everyday applications SEA 11 experts in chemistry, chemistry education, educational psychology, and educational measurement Exploratory Factor Analysis EFA for original sample Kaiser–Meyer–Olkin KMO = 0.92 The bartlett’s test was significant BTS value=3067.45, p0.001, The correlation matrix was significantly different from an identity matrix. The 22 items were factor analyzed and three factors emerged with eigenvalues 1. An oblique rotation direct oblimin All items had pattern coefficients higher than 0.3 51 the three factors were deemed sufficient and conceptually valid in their correspondence to the existing theory. 0.92 for the SCS, .87 for the SPS, and 0.82 for the SEA, indicating satisfactory reliability. Confirmatory Factor Analysis CFA for final sample Analysis of moment structures AMOS version 4, Multiple goodness-of-fit tests including: Normed Fit Index NFI, Comparative Fit Index CFI, and the Root Mean Square Error Approximation RMSEA. Results from the CFA: the three-factor structure fit well to the sample data with all fit indices NFI=0.98; CFI=0.98 indicating a good fit except for RMSEA =0.08, which indicated a reasonable fit. All parameters factor loadings were found to be significant, indicating a significant contribution of each item to the corresponding dimension. Examining item-total correlations indicated that all items in each dimension contributed to the consistency of scores with item- total correlations higher than 0.60. Turning to how the researchers conducted an analysis of data, all researchers used exploratory factor analysis EFA to identify the factor structure of the scale see Tables 3–5. Some of them continued to confirmatory factor analysis CFA to create cross-validating each item and reliability coefficients to estimate each dimension’s internal consistency. In conducting a factor analysis, the first set of factor loadings was obtained by using a method that permits convenient calculation of the loadings. These loadings are called initial or unrotated loadings. Nevertheless, all researchers in these cases used the rotated loading to approximate simple structure. For instance, Uzuntiryaki Aydin 2008 used an oblique N. Suprapto, T.-S. Chang SWUP SC.125 rotation rather than orthogonal rotation. On the other hand, some researchers used varimax as the most popular method for orthogonal rotation. For example, Lin et al. 2014a and Lin et al. 2014b used a varimax rotation to gain principal component extraction. Other ways, some researchers also provided farther content validity evidence. For the purpose of content validation, a group of experts in science view were asked to assess the quality of each item, verify matching of items to the corresponding dimensions, and provide further suggestions. Table 4. Summary of data analysis Lin et al., 2014b. Kind of analysis and the results Dimension Content valid- ation Factor Analysis and T he total percentage of variance extracted Internal consistency from Cronbach’s alpha a Conceptual Understanding 5 items b Higher-order cognitive skills 6 items c Practical work 7 items d Everyday application 8 items e Science communication 6 items EFA Criteria: the retained items should preferably be weighted greater than 0.4. In other words, the items with a factor loading of less than 0.4 were deleted. Principal component extraction with a varimax rotation The Cronbach’s alpha coefficient for each scale of each dimension of the PLSE instrument was calculated The eigenvalues of the five factors from the principal component analysis were all larger than one The total variance explained was 65.60 . Cronbach’s alpha coefficients for the five factors were 0.80; 0.80; 0.90; 0.86; and 0.90 respectively. The overall alpha was .95, indicating that these factors had high internal consistency in measuring the participants’ five dimensions of physics learning self- efficacy. a Higher-order cognitive skills, BLSE-HC 6 items b Everyday application, BLSE, EA 8 items c Science communicatio, BLSE-SC 6 items d Practical work, BLSE-PW 7 items Original five- factor structure The items of Conceptual learning CL’ were eliminated. EFA The KMO value 0.95 and the result of Bartlett’s test x 2 =5329.27, p=0.001 suggested the suitability of conducting factor analysis of the surveyed responses Component analysis with orthogonal varimax rotation to reveal meaningful clusters of factors from the results of the questionnaires. Examined the distribution of the data and revealed the skewness and kurtosis The total variance explained was 63.68 The reliability coefficients Cronbach’s alpha of the factors in the SLSE questionnaire were 0.89, 0.89, 0.91, and 0.89, respectively. For the whole questionnaire, the alpha value achieved 0.96 and represented sufficient reliability to assess the BLSE of university students with biology-related majors. Research on University Student’s self-efficacy scale in science education: A systematic review SWUP SC.126 Table 5. Summary of data analysis Lindstrøm Sharma, 2011. Kind of analysis and the results Dimension Content validation Factor Analysis and T he total percentage of variance extracted Internal consistency from Cronbach’s alpha Only one factor: Physics Self- Efficacy with 5 items Three experienced physics education experts, one of whom is also an expert in self-efficacy and related constructs. They were asked to comment on the validity of the items. Initial trial EFA using the Statistical Package for the Social Sciences SPSS version 15.0. Scree plot, clearly indicated one factor only. Equally there was only one element, factor rotation did not use The five items had factor loadings in the range .694 to 0.821 Confirmatory trial CFA using Amos 7.0 provided evidence for the construct’s validity values in parentheses indicate requirements for validity; χ2 = 2.127, p = 0.831 p.05. Main fit showed a very good model fit: RMSEA =.000 .05 with a 90 confidence interval of [0.000, 0.042]; RMR = 0.009 .05; GFI = 0.998 .95; NFI=0.994 0.95; and CFI = 1.000 0.95. Final checks on the questionnaire were for invariance and stability A questionnaire is said to be invariant if the factor structure for data from different samples from the population is consistent. The factor explained 56 of the variance Cronbach’s α=0.796. 3.2 Implications for the next research For the future research, Figure 2 summarizes the development process of the science self-efficacy questionnaire. The diagram also compatible for developing general self-efficacy and an alternative way how to analyze the data. More ever, there are several criteria for conducting data analysis summarized from four papers were discussed: 1 Exploratory factor analysis was employed to attain the factor structures of the two adopted instruments based on the participants’ responses on the instrument. N. Suprapto, T.-S. Chang SWUP SC.127 2 The validation criteria of exploratory factor analysis: the retained items should preferably be weighted greater than 0.4. In other words, the items with a factor loading of less than 0.4 were deleted. 3 T he condition for factor extraction was based on a combination of Kaiser’s criterion of eigenvalue greater than 1 confirming the intended factor structure at least four-factor loadings over 0.6. 4 The factor explained 56 of the variance values over 50 were acceptable. Figure 2. The process of developing science self-efficacy questionnaire. In addition, Table 6 also described the possibility of dimensions relating to university student self-efficacy in science. In addition, we put two dimensions, namely self-learning strategy and self-assessment as part of the self-efficacy dimension. The main reason is many researchers explored the relationships between university students and their cognitive and metacognitive strategy use and science achievement. In terms of metacognitive, university’s students have self-learning strategy and self-assessment. The field of metacognition in science education is in a state of development and expansion, and that metacognition is increasingly incorporated into research addressing the core aims of science instruction Zohar Barzilai, 2013. Conceptual understanding of science was found to be one of the key objectives of current metacognition research. Their study employed a wide range of instructional practices for fostering learners’ metacognition. The most prominent practice is the use of metacognitive cues and prompts through instruction. It means learning strategy and assessment as part of the course of instruction. Research on University Student’s self-efficacy scale in science education: A systematic review SWUP SC.128 Table 6. The possibility of dimensions relating to university student self-efficacy in science. No Dimension based on literature review Dimension author’s perspective 1 Conceptual Understanding Conceptual Understanding Knowledge comprehension-level skills 2 Higher-order cognitive skills Higher-order thinking skills 3 Practical work Practical work Psychomotor skills 4 Everyday application Everyday application 5 Science communication Science communication 6 Self- Learning Strategy 7 Self-Assessment

4. Conclusion and remarks