As students gain knowledge, their behaviour will improve, fostering positive changes in the teaching and learning process. However, it is challenging to disentangle learning from the educational process. Students must persevere in their academic endeavours, particularly in subjects like chemistry, which are often perceived as formidable. Despite its reputation for difficulty, chemistry holds significant importance in both middle and high school curricula and in many college degree programs (Senocak & Baloglu, 2014). Moreover, (Moreno et al., 2021) emphasize that mastering chemistry demands considerable effort and persistence. Additionally, (Zahro’ & Ismono, 2021) highlight the sequential nature of chemistry and its reliance on computations as contributing factors to students' struggles with the subject. Weak achievement is often associated with a reliance on rote memorization rather than meaningful comprehension. According to (Oladejo et al., 2023), mastery of the subject leads to perceiving its concepts as straightforward.

In secondary schools, the issue of low achievement among chemistry students is crucial and potentially contributes to the challenges associated with grasping chemical concepts (Timilsena et al., 2022). Some students may exhibit overt signs of the effects of a stressful life, such as dread and unwillingness to complete tasks, psychological conflict due to the fear of stopping their studies, feelings of unreliability, and fear of the future, among others. These experiences create a lasting impression on the student's psyche, emphasizing the need for a high degree of conformity to appropriately adapt and deal with difficult life transitions by encouraging academic resilience (Ibrahem Ayasrah & Nawaf Albalawi, 2022). The capacity to overcome obstacles in the classroom, increase motivation and concentration, enhance learning abilities, and adjust to shifting circumstances are all aspects of academic resilience (Suud et al., 2023). Students may have varying personal viewpoints during the learning process, which might alter their behavioural, affective, and cognitive reactions (Alkan & Yucel, 2019).

This goal can be achieved by improving students' learning efficiency and implementing logical processes in chemistry teaching. However, some students can overcome obstacles and difficult circumstances and achieve academic success because they believe learning comes from hard work and commitment, not luck. These students are known as academically resilient students. It is common for students to face difficulties and setbacks in class; if they can adapt and overcome these challenges, it means they have the capacity for academic resilience (Ulya & Gumiandari, 2023). In the educational context, academic resilience is the ability to overcome challenges to excel academically (Cassidy, 2016). Academic resilience is the capacity to deal with academic problems such as academic stress, failure, and boredom (Mahato et al., 2023). On the other hand, resilience describes a person's capacity to overcome obstacles and recover from setbacks (Lakhan et al., 2020). After failed attempts, resilience plays an important role in education and research in the teaching and learning process (Weidlich & Kalz, 2021), and academic resilience is thought to sustain student learning success (Annisa et al., 2023).

Academically gifted children who can maintain their motivation to succeed and excel in class, even when facing difficulties and other situations that could negatively impact their performance, are described as having academic resilience (Mahmudah et al., 2022). Academic success depends on student motivation. Students may only gain academic progress if they can handle challenges, study pressure, and schoolrelated stress (Honra, 2022). If students have good academic resilience, threats can be turned into opportunities for them to grow and learn, fostering positive change (Masten et al., 2021). According to (Rao & Krishnamurthy, 2018), children can be taught academic resilience in the classroom. This is consistent with the explanation by (F. & F, 2015), which states that academic resilience is not merely a quality that a person inherently possesses but rather a dynamic process of adapting to difficulties. It requires actions and behaviours that each individual can adopt and shape.

Low academic resilience in students will hurt their social life, relationships with parents, instructors, and peers, and academic success (Beri & Kumar, 2018). Several elements, such as self-efficacy, self-control, low anxiety, persistence, and planning skills, Evaluating Students' Academic Resilience in Chemistry Learning: Insights from a Rasch Model Analysis influence academic resilience (Beale, 2020). According to (Romano et al., 2021), resilient adolescents appear to overcome stressful situations at school, maintain high motivation levels, and achieve extraordinary performance despite facing difficulties. Resilient students re-engage and persevere through challenging assignments. According to (Reeve et al., 2020), these characteristics predict many positive outcomes, including classroom engagement, enjoyment of school, and overall self-esteem. According to further research, adolescents who demonstrate academic resilience outperform their classmates in protection against severe maladaptation, academic achievement, and school context engagement (Allan et al., 2014)(Fiorilli et al., 2020).

Finding appropriate methods to assess academic resilience in chemistry studies is a challenge. Research on academic resilience shows a strong correlation with individual academic achievement. Currently, there is a significant increase in global research investigating academic resilience (Kumalasari & Akmal, 2020). Research on mathematics resilience is still in its early stages in Indonesia. Although some studies have been conducted in academic institutions, primary and secondary educational settings, and other contexts, the number is limited, and additional studies are still needed to advance understanding in this area.

To provide accurate and objective values, evaluation tools used in educational research need improvement to become better-measuring tools (Muntazhimah, 2019). Research by (Ramdani et al., 2020) developed an Indonesian academic resilience scale with excellent validity and reliability to assess the academic resilience of junior high school students. The scale created can be applied as an evaluation tool when implementing interventions (Nicoll, 2014). To produce more comprehensive instrument information that better meets measurement standards, the junior high school student academic resilience scale will be validated using the Rasch Model ((Briggs, 2019); (Rachman & Napitupulu, 2017); (Winarti & Mubarak, 2019)). This measure of academic resilience is intended to be applicable and based on conditions in Indonesia. The Rasch model, popularized by Danish mathematician Georg Rasch in 1960, is a statistical modelling technique that creates equal interval measurement scales from raw data (Sumintono & Widhiarso, 2015). According to (Bond & Fox, 2007), the Rasch model is a probabilistic response model where the probability of success is based on individual aptitude and item difficulty. By improving the analysis and informing psychometric scale assessments, Rasch's modelling draws on item response theory (Zile-Tamsen, 2017). The advantage of the Rasch Model is that it can identify model inaccuracies and produce more accurate estimates (Taufiq et al., 2021). Rasch Model analysis is not deterministic; it can identify the items being measured more precisely by employing a probabilistic approach to examine the measurement object (Indihadi et al., 2022). The interaction between respondents and question items can be described and measured using the Rasch model (Sumintono et al., 2014). The validity of the instrument was tested using Rasch model analysis. Metrics such as unidimensionality, Wright map analysis, item analysis, participant ability analysis, and instrument analysis are used to assess the quality of the instrument in Rasch model analysis (Muslihin et al., 2022).

To enhance the quality of chemistry teaching and learning for students, the Rasch model is used to measure and validate students' cognitive abilities and the effectiveness of assessment techniques (Chow et al., 2018)(Noben et al., 2021). Meanwhile, some teachers rarely conduct a needs analysis before designing learning tests. However, needs analysis plays a crucial role in learning planning, from designing the syllabus, teaching materials, and class activities to student assessment (Asti Ramadhani). The Rasch model analysis is the best choice for teachers in this situation. Besides interpreting students' knowledge through various Rasch data presentations, studying the structure of questions with varying levels of difficulty can assess students' overall understanding of chemistry. Through Rasch model analysis, students' characteristics and ability levels can be accurately mapped and used to enhance the quality of learning, particularly in the theoretical aspects of the subject matter (Ngadi, 2023). Rasch model analysis provides valuable information about the suitability of an item in measuring students' academic resilience in chemistry learning. By closely examining the results of Rasch model analysis, you can determine not only the reliability of the instrument created but also the patterns of students' responses (Sumintono, 2018)(Chiang, 2015). The development of this academic resilience instrument will provide a more comprehensive and concrete picture of the measurement aspect of the test because the Rasch model involves two basic parameters: (a) student ability and (b) the level of difficulty of questions or the ability to ask questions (Khairani & Bin Abd. Razak, 2015). Therefore, Rasch modelling is important in chemistry learning because it can be applied across various scientific disciplines and student grade levels. This research tests measurements of academic resilience based on Indonesian students' conditions, with few studies specifically focusing on chemistry learning. This encourages researchers to create measuring tools, such as instruments for assessing students' academic resilience in chemistry while maximizing their utilization during the learning process. Evaluation instruments in educational research must be developed into robust and high-quality measuring tools to yield objective and accurate results (Muntazhimah, 2019). As a result, teachers gain valuable information to enhance the quality of teaching and evaluate chemistry learning. This is supported by research (Cheung et al., 2011), which emphasizes the importance of conducting student diagnosis and learning processes as strategies for designing effective teaching and improving students' understanding of chemistry. Instrument quality measurements in this research include unidimensionality, Wright map analysis, item analysis, participant ability analysis, and instrument analysis (Planinic et al., 2019). According to (Grift et al., 2019), the Rasch model can be a useful tool for classroom assessment. Rasch is a technique for assessing student aptitude, and it can also be used to evaluate the quality of questions presented in a statistical data format (Almubarak et al., 2023).

The conceptual framework developed for this study, depicted in Figure 1, aims to provide more precise direction and guidance. To better align with the culture and characteristics of Indonesian students, the academic resilience questionnaire has been modified. This study intends to examine the validity and reliability of a new instrument designed to measure students' academic resilience in chemistry learning.

This study employed a quantitative technique and a survey design. Through survey research, information from a sample of people is obtained, allowing the generalization of knowledge about the population (Gul, 2023). This study aims to examine the academic resilience instrument, which involves consultation with three experts ((Cassidy, 2016);(Idris et al., 2019); (Ramirez-Granizo et al., 2020)). The validation procedure for this academic resilience tool uses the Rasch Model. The Rasch Model can be used to (a) facilitate the development of instruments that produce valuable data, (b) produce data that can be used confidently for parametric and descriptive statistics, and (c) generate outcome measures that provide practitioners and researchers in school psychology with clinically relevant guidance (Boone & Noltemeyer, 2017).

The study's population consisted of tenth and eleventh-grade students from 13 different schools. The survey had 817 participants from Pekanbaru, Indonesia, selected using simple random sampling techniques. This method was employed to minimize data bias, as random sampling involves randomly selecting members of the population, ensuring every individual has an equal opportunity to be chosen. Random sampling is considered the most straightforward approach for data collection.

A Google Forms-based academic resilience questionnaire was used to collect data for this study. Google Forms provides more than just an assessment tool; it is a tried-and-true method for gathering data. Additionally, teachers and students can communicate via Google Forms (Rodriguez, 2018). Creating an online survey with Google Forms is quick and straightforward, and the results are stored in an online spreadsheet (Ramaraj, 2019). The questionnaire consists of three components comprising twelve statement items, specifically designed to cater to the requirements of Indonesian students enrolled in chemistry programs. The research instruments are accessible online, facilitating the data collection process. This grid of academic resilience instruments is intended for students gathering data related to chemistry. Table 1 illustrates it as follows.

Rasch model data analysis was performed using Winstep software, designed by (Linacre, 2011). Respondents and items interact simultaneously, according to the Rasch model. Instead of a raw score, the Rasch model displays a logit value representing the likelihood of selecting an item among respondents. This logit value is derived after applying the logarithmic function to the odds ratio value of the item, transforming the raw score into the item's logit value. According to (Sumintono et al., 2014), The odds ratio value is a probability number that indicates the degree of respondents' agreement with an item compared to respondents who disagree. Our evaluation of the item becomes more objective when we use its logit value, as it converts the ordinal raw score into ratio data that meets all integer requirements.

The data used in the Rasch model is the probability score (P), or the comparison between the number of questions and the correct answers. The following formula is then used to translate the probability score into an odds ratio value:

Next, we use the following formula to obtain the logit value by applying the logarithm function (Sumintono et al., 2014)

This value is called the logit W-score or measurement value. The logit value has been scaled and can be used for various analyses.

Survey questions with ordinal or polytomous responses can also be calibrated using Rasch analysis. The Polychotomous Rasch Model is used in this type of problem. The Rating Scale Model (RSM) and the Partial Credit Model (PCM) constitute the Polychotomous Rasch Model. PCM, a refinement of Rasch analysis, is employed to analyze questions with polytomous or ordinal-type responses. Because PCM can handle questions with multiple categories of reactions, this method is highly relevant for assessing questions in education and other fields (Bond & Fox, 2007). Research using PCM for question analysis was conducted by (Wahyuningsih, 2021).

The PCM model is used to analyze cases involving polytomous responses based on Rasch's analysis. The following formula represents the fundamental Rasch analysis model:

The elements in the formula above can be explained as follows. For the PCM model, it has the following formula.

An explanation of each component in the PCM formula is as follows.

Pr (Xni=x) : probability of the nth participant with ability θ to answer to get a score x on the ith question

x : participant's score

θ_n : ability of the nth participant

δ_ik : level of difficulty for stage k in question item i (Wu et al., 2016).

A rating scale validity analysis test aims to determine whether respondents find the rating alternatives confusing. Using Likert ratings as an example, five options are provided ranging from STS (Strongly to Disagree Strongly), TS (Tend to Disagree), N (Neutral), S (Tend to Agree), to SS (Strongly Agree), all representing polytomous data. The instrument's ranking assumptions can be verified through Rasch model analysis. The Rating (Half Credit) scale test is utilized in the Winsteps program to measure rating scales. The observed average and Andrich thresholds, which indicate the precision of the options provided to respondents, are displayed as results.

Unidimensionality is a crucial metric for assessing whether a new instrument can capture the essence of the target, in this case, the academic resilience of chemistry students. Principal Component Analysis (PCA) of residuals is used in Rasch model analysis to determine how effectively various items assess what needs to be measured (Misbach & Sumintono, 2014).

Using various measurement models, including Infit, Outfit, Mean-Square Fit Statistics, and Standard Fit Statistics (ZSTD), the Rasch model is employed to examine academic resilience tests. According to (Linacre, 2002), two statistics determine whether data are suitable for the Rasch model: Outfit (outlier-sensitive or information-weighted fit) and Infit (inlier-sensitive or information-weighted fit). These statistics are typically presented in mean squared (MNSQ) and z-standardized (ZSTD) formats. According to (Bond & Fox, 2007), ZSTD is a standardized form of MNSQ that adjusts for sample size. MNSQ, on the other hand, represents the average of the squared residuals for an item.

Therefore, the Winsteps Rasch program output in the form of Outfit Mean Square (MNSQ) data needs to be examined to determine whether an item or respondent (person) fits or misfits the Rasch model in this research. Because it is not influenced by sample size, the MNSQ statistic was selected.

The model suitability for measurement can be evaluated using the following general guidelines, according to (Linacre, 2002): MNSQ > 2.0 indicates harm to the measurement system; 1.5 < MNSQ ≤ 2.0 indicates it has no significance for measurement; 0.5 ≤ MNSQ ≤ 1.5 indicates useful for measurement, and MNSQ < 0.5 means it is not helpful for measurement even though it does not harm the measurement system. Misfit items in a measurement tool should be avoided since they don't add much to the test score's dependability (Zubairi & Kassim, 2006). If an item is found to misfit the Rasch model, it may indicate several potential issues with its construction or validity. According to (Zubairi & Kassim, 2006), misfitting items could have Analysis problematic or flawed construction, such as low discriminatory power, meaning they fail to distinguish between different levels of the trait being measured effectively. Alternatively, misfitting items may have questionable validity, suggesting that they measure an ability or attribute unrelated to the assessed intended construct. Identifying and addressing misfitting items is crucial in ensuring the accuracy and reliability of measurement instruments, particularly in academic resilience assessments where precise and valid measurement is essential for meaningful educational research and practice (Reise, 1990).

ZSTD is also a t-test for a hypothesis, which inquires whether the data matches the model. As a result, the predicted value of z is almost equal to zero. The z value has a mean practically equal to 0 and a standard deviation of 1 when the observed data agrees with the model. Giant (z > +2) or low (z < -2) ZSTD values suggest that the item does not fit into the predicted model. The standardized z value (ZSTD) of outfit and infit can be negative or positive. ZSTD values that are negative signify less variation in the model. All participants could answer the question correctly, but all subjects with low ability answered it erroneously. This approach to the responses was close to the Guttman-style response string model. A positive value, on the other hand, denotes that answer variations exceed model variations. Reactions are erratic and unforeseen (Bond & Fox, 2015).

According to (Boone et al., 2014), the criteria applied are the same ones used to confirm that the question items are appropriate.

This suggests that question items of poor quality should be reconsidered or replaced if they fail to meet these two criteria. The sample size significantly influences the applicability of these criteria rather than the consistency of item difficulty alone. Errors in answer keys, careless responses from a large number of participants, and questions with low discriminatory power can all contribute to an item's inadequacy. It's essential to recognize that the ZSTD value is heavily influenced by sample size; it tends to exceed three with larger samples (>500). Therefore, some experts caution against relying solely on the ZSTD criterion when the sample size is substantial (Sumintono & Widhiarso, 2015).

Validity and reliability are critical considerations when developing new research tools. Using the Rasch model to analyze academic resilience instruments ensures accurate, impartial, and consistent data by capturing interactions between respondents and statement items. The study's findings confirm the validity and reliability of the developed academic resilience instrument. The instrument demonstrates robust performance with an item reliability value of 0.99, indicating exceptional quality, and a person reliability value of 0.78, reflecting good consistency among respondents' responses. Specifically designed to assess students' academic resilience in chemistry education, this novel resilience scale shows promising utility. However, further research is recommended, significantly if extending its application beyond the Indonesian context or into higher education.