COMPARING MODELS OF ACCREDITATION AND ACADEMIC PERFORMANCE RELATIONSHIP USING GENERALIZED STRUCTURAL COMPONENT ANALYSIS

The National Education Standards Board in Indonesia is an independent and professional institution which maintain and control the quality of education. The National Examination is one of indicator which can be used as a basis for evaluating quality of education. National Accreditation Board for Schools conducts assessment to the schools on fulfillment of the established standards through accreditation process. There are several theoretical models of relationship between 8 national education standards for describing causality each other. The objective of this study are (1) to compare and determine the best model of the relationship between eight standards using generalized structured component analysis; and (2) to evaluate validity of indicators of accreditation instrument. It has been concluded that the model published by the Ministry of Education and Culture (2017) was the best model. AVE and Cronbach’s alpha showed that score on Mathematics, Science, English and Indonesian Language are important indicators for academic performance. Critical ratio and variance inflation factor showed that there are 13 of 124 indicators of accreditation instrument are not valid. Analysis of structural model showed that school management has a big influence on standard of teachers and education staff. In addition, curriculum, standard of competency, standard of assessment and standard of process have direct influences to academic achievement.


Introduction
Education influences various aspects of life. With rapid changing of technologies in almost all aspect of life, education system needs to prepare students to think using higher order skills. The goal of education is not only for expanding the access, but also for improving the quality of education through developing curriculum, learning process, infrastructure, management, teachers etc. Although since early 2000s, education spending in Indonesia has increased dramatically, but the quality of education is still left behind from many other counties in the world.
The National Education Standards Board (BSNP) as an independent and professional institution in Indonesia which has the task to maintain and control the quality of education. BSNP prepares and develops the National Education Standards (NES) as a basis for planning, implementing, and monitoring education in order to realize quality national education.
Based on regulation, definition of NES is a minimum criteria that has to be fulfilled by schools. NES consists of eight standards, namely standard of content (SI), standard of process (SPR), standard of competency (SKL), standard of teachers and education staff (SPT), standard of infrastructure (SSP), standard of management (SPL), standard of budget (SB), and standard of assessment (SPN).
The National Accreditation Board for Schools and Madrasah (BAN-S/M) conducts assessment to the schools and madrasah on fulfillment of the established standards by BSNP through accreditation process.
Theoretically, fulfillment of SNP can be characterized by relationship between accreditation and achievement of national examination (NE), because the national examination is one of good indicators which can be used as a basis for evaluating of NES achievements. Schools with good accreditation are expected to have a good NE score.
Eight standards in the NES are used as a basis for developing of accreditation instrument. The eight NES are latent variables that cannot be measured directly. The eight SNP assessments were measured by 124 indicators for junior secondary education level. There are several theories that explain the relationship between eight SNP that have been published, namely the Ministry of National Education and Ministry of Religion  Hwang and Takane (2004) developed the GSCA method to overcome limitations on CBSEM and PLSPM in which it does not require multivariate normal distribution but it has overall goodness of fit. This study aims to compare and determine the best model of the three relationship models of 8 education standards and academic achievement in Indonesia using GSCA method.

Data
The study uses accreditation and national assessment data of junior secondary schools in Indonesia consisting of 2069 data in 2018. Accreditation data consists of 124 observable indicators with scale from 0 to 4 from 8 latent variables, while national examination data consists of exam score of four subjects, namely English (ING), Indonesian Language (BIN), Mathematics (MTK) and Natural Sciences (IPA). Table 1 describes list of latent variables and their corresponding number of each observable indicators.

The Steps of Data Analysis
The data analysis is conducted with the following steps: 1) Carrying out data exploration by looking at correlation and comparing accreditation status with average of exam score.

Data Exploration
The number of observations of accreditation and national examination is 2069 junior education schools consisting of 877 public schools, 867 private schools, 62 public madrasahs and 263 private madrasahs. The overall percentage of school accredited A is 58.48%, accredited B is 35.23%, accredited C is 6.09% and not accredited is 0.19%. Based on Table 2, the correlation coefficient between the eight standards and national score shows a fairly high and positive. Figure 3 also shows that schools with better accreditation have higher average exam score in each field of study. Setiawan et al (2018) also concluded that there is a relationship between accreditation status and national examination.

Evaluation of Measurement Models
Evaluation of measurement model with a reflective indicator variable was carried out by assessing the convergent validity, discriminant validity and composite reliability. Based on Table 3, the value of the loading factor for model 1, model 2 and model 3 in each indicator shows a value greater than 0.70 and significant at the significant level of 5%. It can be concluded that the indicators for the latent variable of PA have good convergent validity (Hwang and Takane, 2004). Cronbach's alpha shows a value greater than 0.70 which means that the indicator for the latent variable PA has good composite reliability. While the √ value obtained for model 1, model 2 and model 3 is 0.951. If the √ value is compared with the correlation between PA and other latent variables in Table 4, then the √ value is greater than the correlation value between PA and other latent variables. This shows that the indicator for the latent variable PA has good discriminant validity.
Evaluation of the measurement model with a formative indicator variable is carried out by assessing the significance of its weight. The evaluation results of model 1 show that there are 14 indicators are not significant, namely 13, 17, 38, 39, 46, 51, 55, 57, 59, 74, 75, 76, 80, and 108 at the significance level of 5%.  based on the VIF value with a recommended value of less than 10. The results of multicollinearity test on each indicators give VIF value less than 10 so that all indicators have met the assumptions of multicollinearity.

Evaluation of Structural Model
The structural model is evaluated by identify the parameter coefficient and the significance of these parameters. Figure 4 presents a structural model based on the Ministry of National Education and Ministry of Religion (2010) with R-square and coefficients for each path. In the model, the relationship between SI and PA with a path coefficient of 0.012 is not significant. This shows that SI has relationship between PA but does not significantly affect at the significance level of 5%. Figure 5 presents a structural model based on the Ministry of Education and Culture (2012) with the R-square and coefficient of each path. In this model, the relationship between SI and PA and the relationship between SPT and SPN are not significant. This shows that SI has relationship between PA and SPT has relationship between SPN but do not significantly affect at the significance level of 5%. Meanwhile, Figure 6 presents a structural model based on Ministry of National Education (2017) with R-square and coefficients for each path. In the model, the relationship between SI and PA with a path coefficient of 0.018 is not significant. This shows that SI has relationship between PA but does not significantly affect at the significance level of 5%.
The latent variables that give direct influence on PA in both models are SKL, SI, SPR and SPN, but the direct influence of SI on PA is not significant. The greatest influence on PA is the SKL. The R-square value of each latent variable in model ranged from 0.200 -0.720. This means that the minimum variant that can be explained in both model is 20%.

Overall Goodness of Fit
The FIT values generated for model 1, model 2, and model 3 are 0.603, 0.623 and 0.630. Based on the FIT value, it can be interpreted that the total diversity of all variables that can be explained by each model is 60.3%, 62.3%, and 63%. Based on the percentage variety value, it can be concluded that the model published by Ministry of National Education (2017) better describes the variety of data than the other two models.

Conclusion and Recommendations
In this study, it can be concluded that the model published by the Ministry of Education and Culture (2012) is the best model to describe relationship between 8 national education standards. This model indicates 13 indicators are not valid, namely items 13, 38, 39, 46, 51, 55, 57, 59, 72, 74, 75, 80, and 108. Standard of competency (SKL), standard of process (SPR), and standard of assessment (SPN) have a significant effect on academic performance (PA).