The Development of Testlet Assessment Instrument Model Integrated with E-ujian Website to Measure the Higher-Order Thinking Skills

: This research & development research aimed to determine the feasibility and characteristics of the Testlet model assessment instrument integrated with the E-ujian website in measuring the Higher-order Thinking Skills (HOTS) of junior high school students. The researchers employed the 4D development model (define, design, develop and disseminate). However, the researchers conduct the development up to the develop stage. The research was conducted at a junior high school with 42 students as respondents. The researchers collected the data through test techniques, questionnaires, interviews, observation, and documentation. This development research utilized the descriptive-quantitative data analysis method through characteristic tests, teacher and students’ questionnaire responses analysis, and analysis of students' HOTS profiles. The students’ HOTS profiles based on the assessment instrument were divided into 53 percent of students in the poor HOTS category, 31 percent of students in the low HOTS category, and 16 percent of students in the high HOTS category. Concluded that the assessment instrument developed was very feasible, with a percentage of 94 percent. The assessment instrument developed had the characteristics of valid, reliable, discriminatory, disproportionate, relevant, representative, impractical, and specific. The Testlet model assessment instrument development can be used as an alternative for the distance learning assessment during the Covid-19 pandemic.


INTRODUCTION
Competency learning as a part of the 2013 curriculum contains authentic learning and assessment processes so that students can achieve attitude, knowledge, and skill competencies (Asrizal et al., 2018;Nurtanto et al., 2021). Learning objectives will be achieved if the learning process and assessment show that all learning competencies are achieved and mastered by students. Masitoh & Aedi, (2020) explain that the knowledge competence (cognitive domain), especially in the content standards of the 2013 curriculum, is designed so that students can have high-order thinking skills (HOTS). Therefore, the assessment must measure students' HOTS appropriately.
The low level of higher order thinking skills (HOTS) can be developed through the learning process and learning assessment (Istiyono et al., 2018;Widana, 2018). The standard for assessing student learning outcomes must focus on HOTS. HOTS mastery enables students to develop themselves in deciding, assessing, and solving problems appropriately. Prayitno & Nofiana, (2020) explain that there are five skills that makeup HOTS: (1) critical thinking skills, (2) creative thinking skills, (3) problemsolving skills, (4) decision-making skills, and (5) argumentative skills.
Higher-order thinking skills (HOTS) can make students connect, manipulate, and transform their knowledge and experience to think critically and creatively (Arif & Yuhdi, 2020;Arnellis et al., 2020;Halim et al., 2018;Nisa et al., 2018). One of the benefits of using HOTS in learning is that students' information or knowledge will be stored longer than just using Lower-Order Thinking Skills (LOTS). Besides, HOTS can also make students develop themselves in deciding and solving problems appropriately. The 2015 TIMSS (Trends in International Mathematics and Science Study) showed that Indonesia is ranked 44th out of 49 countries globally. This result shows that the students' HOTS level was poor (Syamsul Hadi & Novaliyosi, 2019). Students often work on questions that tend to LOTS. Low HOTS can be caused by the model or learning method that cannot develop students' HOTS (Retnawati et al., 2018;Tan & Halili, 2015). Also, according to (Rahayuningsih & Jayanti, 2019), the low HOTS can be caused by students who are less accustomed to working on HOTSbased questions.
Assessment instruments are essential in learning. The quality of the assessment instrument directly affects the accuracy of the student's competency achievement status (Dharmawati et al., 2016;Qi & Mitchell, 2012). It can be concluded that an appropriate assessment instrument is needed to measure the HOTS of junior high school students appropriately.
One of the assessment instruments that can be used to measure students' HOTS is a Testlet. A Testlet is a collection of items containing information on the same topic where the items have been considered a single unit that shares the same problem contest (Nova et al., 2016). Lutviana et al., (2019) explain that the Testlet model questions have more practicality than the description form test because the scoring can be done objectively and is polytomic. Testlet is an assessment instrument that can combine the advantages of multiple-choices questions and descriptions (Safitri, 2020;Shiell & Slepkov, 2015;Slepkov, 2013;Wahyuni et al., 2015).
The items in the Testlet model test consist of main questions and supporting questions. The questions are made to provide information to other supporting questions. The supporting items are made to have a level of completion of the main problem so that it is expected to help teachers diagnose students' learning difficulties (Muna et al., 2017). Beside diagnosing learning difficulties, Testlet can also measure students' skills deeper (Fahmina et al., 2019). (Shiell & Slepkov, 2015;Slepkov & Shiell, 2014) explain that the correct answer to each question can give students an idea of the whole, partly or not, regarding mastery of the following questions. The scoring method used in the Testlet is the Graded Response Model (GRM).
Fahmina et al., (2019) utilize a computerized Testlet to measure the literacy skills of high school students. Computerized assessment can also be done using the help of websites. One of the websites that provide services to carry out online assessments or assessments is E-ujian. E-ujian facilitates teachers to carry out online assessments. It can be accessed via a computer or smartphone to assess anywhere. The tracking feature through the camera and sharing sessions in the exam implementation can help teachers monitor and minimize cheating committed by students during examinations.
The development of Testlet model assessment instruments has been widely carried out, one of which is the development carried out by (Nova et al., 2016). They have succeeded in developing an assessment instrument on temperature and heat material for the tenth-grade high school students. Other research has also been carried out by (Muna et al., 2017), who succeeded in developing a Testlet model test instrument to detect students' learning difficulties on buffer solutions topic. (Damayanti, 2017) developed a Testlet model assessment instrument to measure higher-order thinking skills of high school students on electrochemical material. The higher-order thinking skills indicators were the skills to analyze, evaluate, create, think critically, and think logically.
Based on the description, research gaps have not been investigated further. One of the gaps is further research on the Testlet model assessment instrument to measure the students' HOTS with different indicators. Therefore, the solution to this problem is to develop an integrated Testlet model assessment instrument on the E-ujian website to measure the higher-order thinking skills of junior high school students.

METHOD
This research is development research with a 4D development model. Thiagarajan's 4D development model consists of four stages: define, design, develop, and disseminate (Thiagarajan et al., 1974;Utomo & Kustijono, 2015). However, the researchers conducted the research only up to the develop stage. This development research involved lecturers and science teachers who were tasked to validate the developed product and the research subject (the seventhgrade junior high school students). The number of students who became the research subject consisted of 42 students.
Two tests were carried out at the develop stage on the developed products.
First is the small-scale trial to determine the students' and teachers' responses to the Testlet model assessment instrument. There were ten students and four teachers involved in the small-scale trial. Second, a large-scale trial was conducted on 32 students to determine the characteristics of the assessment instrument developed. At the same time, the large-scale trial data was also used to analyze the students' HOTS profiles.
Instruments used to collect data in this study consisted of expert validation sheets, teacher response questionnaires, student response questionnaires, and question scripts. The experts involved were material and assessment experts. Expert validation sheets were used to assess product feasibility. Students' and teachers' questionnaire responses were used to determining the product's characteristics. The characteristics were relevance, representativeness, practicality, and specificity. On the other hand, the question text was used to determine the product's general characteristics (valid, reliability, discriminatory, and proportional) as an assessment instrument. The Testlet model was developed based on five HOTS aspects. The HOTS indicators can be seen in Table 1. The data used to analyze the HOTS profile of students was obtained from the results of the analysis of questions with cognitive levels C4-C6 containing indicators of the HOTS aspect. The questions developed in the Testlet model assessment instrument consisted of 15 main questions, where each main question was integrated with three supporting questions. The questions developed were questions with cognitive levels C1-C6. The questions with the HOTS cognitive level were contained in the main questions numbered 1, 3, 5, 6, 7, 10, 11, 12, and 14. Each main question was integrated with three supporting items so that there were 27 total questions used for students' HOTS analysis.
The feasibility analysis of the instrument was carried out through the percentage descriptive method. The feasibility value was obtained from the expert validation questionnaire sheet. The feasibility value obtained was then interpreted into four criteria, namely (1) highly feasible if the percentage is 81.25 < x 100.00, (2) feasible if the percentage is 62.50 < x 81.25, and (3) not feasible if the percentage is 43.75 < x 62.25, and (4) highly not feasible if the percentage is 25.00 x 43.75. The assessment instrument for the integrated Testlet model on the Eujian website can be feasible if it has a percentage higher than 62.50 in the feasible and highly feasible criteria.
The validation was carried out by five material experts and five assessment experts. The results of expert validation were analyzed using the Aiken formula to obtain the content validity values of the items from the assessment instrument. Each item is said to be valid if the validity test results show that Vcount is greater than Vtable with a significant level of 5 % (0.78 for assessment and material experts).
The reliability of the developed Testlet model assessment instrument was analyzed using the Cronbach Alpha method. In this method, the value of the reliability coefficient alpha (r11) is compared with the value of the rtable with a 5 % significance level. If the rcount is higher than rtable, the question is declared reliable.
The characteristics of discrimination can be analyzed by performing a discriminatory test. The discriminatory power is the skills of a question to distinguish between high-skills students and low-skills students (Arikunto, 2013). The discriminatory power test used was a discriminatory power test to assess question descriptions. Discriminatory power was obtained by comparing the difference between the upper and lower averages with the maximum score of the Tetslet model assessment instrument being tested.
The researchers interpreted the discriminatory power criteria from the results of the calculation of the discriminatory power of the items. There were four criteria used in the discriminatory power test. The question is accepted if the discriminatory power test is in the interval of 0.40 ≤ D ≤ 1.00. The question is accepted but needs to be corrected if the discriminatory power test is in the interval of 0.30 ≤ D ≤ 0.40, while the question needs to be corrected if it is at the interval of 0.20 ≤ D ≤ 0.40. The question cannot be used if the discriminatory value is 0.00 ≤ D ≤ 0.20.

Proportional
analysis was performed by conducting a test of item difficulty. Rusilowati et al. (2016) state that the difficulty level test can be done by comparing the average score with the maximum score obtained on each item. The results of the difficulty level were interpreted against the existing criteria. There were three criteria for the level of difficulty: (1) Difficult if the coefficient of difficulty level is in the interval of 0.00 ≤ ID ≤ 0.30, (2) Medium if the coefficient of difficulty level is in the interval of 0.30 ≤ ID ≤ 0.70, and (3) Easy if the coefficient of difficulty level is in the interval of 0.70 ≤ ID ≤ 1.00. Items that are good to use are those with the criteria of the interval of 0.70 ≤ ID ≤ 1.00 or having moderate criteria.
Questionnaire analysis of student and teacher responses was carried out to analyze the relevant, representative, practical characteristics and specifications of the assessment instrument that had been developed. The answer to each question or statement is calculated as the Statement of Approval Level (SAL) using the representative descriptive method, comparing the total scores obtained with the total number of ideal scores.
The researchers interpreted the questionnaire assessment criteria from the results of the calculation of the level of approval of the statement. There are four criteria used, namely (1) excellent when the SAL is in the interval of 75.00 < SAL ≤ 100.00, (2) high when the SAL is in the interval of 50.00 < SAL ≤ 75.00, and (3) low when it is in the interval of 25.00 < SAL ≤ 50.00, and (4) poor when the SAL is in the interval of 0.00 ≤ SAL ≤ 25.00.
HOTS analysis was carried out using the GRM method. The Graded Response Model (GRM) is a scoring model that applies a grading system to assess the answers to questions (Mateucci & Stracqualursi, 2006;Morse et al., 2012). GRM is one of the models developed to handle scoring on polytomic items (LaHuis et al., 2011;Momani, 2017). Based on the GRM scoring method, students will get a maximum score of 3 if they can complete stages 1, 2, and 3 correctly. On the other hand, they will get a score of 0 if they cannot complete the first stage correctly. The results of the large-scale trial were used to determine the students' HOTS for each indicator. The percentage of students' HOTS was then calculated using the following mathematical equation.  Table 2.

RESULT AND DISCUSSION The Feasibility of Testlet Assessment Instruments Model Integrated with Eujian Website
The feasibility of the product is determined by calculating the total score obtained from the assessment and media experts' validation using the percentage descriptive technique. The feasibility percentage from the assessment experts was 94 %, while the feasibility percentage from the material experts was 95 %, both in the highest feasible category. The experts stated that the Testlet model assessment instrument integrated with the E-ujian website to measure the higherorder thinking skills of junior high school students was highly feasible to be used in the learning process with an average percentage of 94.5 %.
The aspects assessed by the assessment expert were the content and construct aspects. The percentage obtained for both aspects was 95%, which was declared highly feasible. Each aspect was assessed through indicators. In the content aspect, the indicator that got the highest percentage was "The relationship between competencies, Competency Achievement Indicators (GPA), questions, questions, and Bloom's Taxonomy level," with a percentage of 100%. The indicator that got the lowest percentage was the '' The suitability for measuring HOTS'' with a percentage of 90%.
The construct aspects of the assessment expert validation were assessed through four indicators: (1) ease of understanding, (2) systematic, (3) practicality, and (4) use of proportional typeface and font size. The indicator that received the highest percentage was ''easy to understand'' with a percentage of 100 %. The other three indicators get a percentage of 95% with highly feasible criteria. The assessment experts reviewed four aspects: competence, material quality, display, and language. The average percentage of all aspects assessed was 94%, with highly feasible criteria. The percentage of display indicators was 100% in the highest feasible category. Meanwhile, the lowest percentage (90%) was found in the indicator of the relationship between competencies, Competency Achievement Indicators (GPA), questions, and questions and material indicators. The language indicators contained the language assessment aspect.
Besides producing feasibility data from the assessment instrument at the validation stage, the Testlet model integrated with the E-ujian website also obtained suggestions and input. The suggestions and input were used for improvement. Improvements were made to improve the quality in terms of assessment and material. One of the improvements based on assessment experts' input related to the question grid. Improvements were made by adding a cognitive dimension mapping column. Improvements to the grid were also carried out by removing the HOTS indicator on questions with LOTS and MOTS cognitive levels. The HOTS indicator on the question was replaced with the Competency Achievement Indicator (GPA). Other improvements were also made based on the advice of the material experts. The design improvements were also carried out by improving the writing and layout of images, discourse on the main questions, supporting questions, and answer choices. Improvements to the questions on the Eujian website page can be seen in Figure  1.

Figure 1. Improvement on the E-ujian Website
Item number three in Figure 1 was improved by increasing the size of the image to be easily observed by students. These improvements were made based on the advice of assessment experts and materials experts.

Characteristics of Assessment Instruments Testlet Model Integrated E-ujian Website Valid
The results of the content validity analysis stated that the developed assessment instrument had high validity. (Arikunto, 2013;Hayashi et al., 2019) Explain that a test can be enriched as valid if it can measure what it wants to measure. All of the developed questions (45 items) can be declared valid. Five assessors determined the item validity with a significant level of 5%. The results showed that Vcount was higher than Vtable (0.87).

Reliability
The researchers analyzed the reliability using the Cronbach Alpha formula assisted by the Microsoft Excel 2016 application. Based on the results of the analysis, the reliability value (r11) was 0.70 for 32 students in the large-scale trial (class VII E). (Rusilowati et al., 2016) state that the reliability of the instrument is high if the reliability coefficient (r11) is 0.6 (r lower than 0.8). Based on the analysis results, the Testlet model assessment instrument can be trusted and provides the same results if used on different occasions, such as on different subjects, conditions and places, and times.

Discriminatory
Discriminatory characteristics can be known through the analysis of discriminatory tests. The data used to analyze discriminatory power was obtained at the small-scale trial stage. The average value of discriminating power for all items developed was 0.40, which was included in the high category. The results of the analysis of the discriminatory test showed that from the 15 main questions developed. Three main questions had sufficient criteria. These items are the main questions number 1, 7, and 10.
The difficulty level for all the main items developed was 0.35 in the medium criteria. The questions used to measure students' HOTS profiles had an average difficulty level of 0.29 in the difficult criteria. HOTS questions are found in the main question numbers 1, 3, 5, 6, 7, 10, 11, 12, and 14. The difficult question criteria were obtained because students were unfamiliar with Testlet-type questions and the limited learning time during distance learning, so the material provided was not studied thoroughly. (Lichtenstein & Fischhoff, 1977;Maulida et al., 2015) State that the category of item difficulty level depends on the skills of the research subject and the general skills of all research subjects (not only individuals).

Relevant
The Testlet model assessment instrument integrated with the E-ujian website to measure students' HOTS was relevant, with an average percentage of 93.65 % for all indicators measured. The indicators measured in the relevant characteristics are (1) the substance of the questions refers to the scope of the competency achievement indicators, and (2) the questions instruments can interpret the cognitive domain. (Safitri, 2020) explains that preparing an assessment instrument that follows the competence will allow the alignment of the instrument form, material, and time allocation required with the depth of the material.

Representative
The questionnaire analysis results showed that the assessment instrument developed represented all the material taught. The percentage obtained was within the excellent criteria (96% for all indicators at all stages). Recapitulation of the questionnaire responses results can be seen in Table 4.12. The results of this analysis are also supported by the analysis of the material experts at the validation stage. The quality of the material in the material expert validation sheet contained the accuracy and truth indicators.

Impractical
The practicality based on the validation of the assessment experts showed that the developed assessment instrument was declared highly feasible with a validation percentage of 95 %. The assessment instrument was easy to grade, making it easier for teachers to correct so that an objective and fast assessment is produced and minimizes student cheating during the test. Students also get a discussion at the end of the assessment session. The weakness of the developed assessment instrument was that it was a pay-to-use platform, so it is not appropriate if used for formative assessments, such as daily tests. Besides, the camera feature caused slow website access in large quantities and numbers. At the same time, the website often experienced a decline in function. Also, students at junior high schools were not familiar with the website, which affected the implementation of the test instrument.
The Testlet model assessment instrument integrated with the E-ujian website is considered practical if it is easy to use. This aligns with (Arifin, 2012) statement, which explains that practical means are easy to use. The practicality of an assessment instrument is seen not only from the manufacturer's side but also by people who want to use the instrument. Based on the questionnaire results, the developed product was declared impractical because it has many weaknesses compared to its advantages.

Specific
The questionnaire analysis of student and teacher responses shows that the Testlet model assessment instrument was specific within excellent criteria. The average percentage was 91%. The assessment instrument was specific because it had questions that focused on the science object material and observations.
Another specific characteristic of the Testlet model assessment instrument is its usability and characteristics that distinguish it from other instruments. The Testlet model assessment instrument developed can measure junior high school students' HOTS.

Students' HOTS Profile
The students' HOTS profile data was obtained from the implementation results in the large-scale trial. The measured HOTS aspects consisted of five skills, namely: (1) critical thinking skills, (2) creative thinking skills, (3) problemsolving skills, (4) decision-making skills, and (5) argumentative skills. A recapitulation of the number of items for each HOTS aspect indicator can be seen in Table 1. The Testlet model questions developed in this study were adapted from (Shiell & Slepkov, 2015), an integrated set of questions consisting of three or more items in the form of multiple tiered choices. The items in the Testlet model test consisted of main questions and supporting questions. The questions were made to provide information to other Every main question contains three supporting questions that are related to each other. In the supporting questions, some indicators can measure the HOTS aspects of students. Item 1.1 measures the HOTS aspect in the form of the skills to argue with the indicator to be measured, namely Data. Item 1.2 measures the HOTS aspect of critical thinking skills with indicators in the form of combining information. Item 1.3 measures the aspect of problem-solving skills with indicators of compiling problem-solving. Therefore, there are two to three aspects of HOTS. The data from the large-scale trial analysis on 32 seventh-grade junior high school students related to the students' HOTS profiles are presented in Figure 2. The students' HOTS profiles were divided into three categories. There were 53 % of students in the very low HOTS category, 31 % of students in the low HOTS category, and 16% in the high HOTS category. Based on the results of data analysis on proportional characteristics, the questions developed were in the difficult category.

Achievement of HOTS Aspects for Students with High HOTS
There were 16 % students out of 32 students belonging to the high HOTS category. The students in this category were familiar with the level of difficulty of the questions on the Testlet model assessment instrument. This fact is supported by the results of student questionnaire analysis in the very good HOTS category of 75 %. Questionnaire items for student responses contain statements stating that students are familiar with the difficulty of the questions being worked on. The high HOTS of students is also due to the understanding of concepts, application and use, utilization, and selection of operations that have reached the excellent category. This fact is coherent with the results of the questionnaire analysis, which shows the percentage of achievement in that aspect is 75 % and 70 %. The highest aspect is the skills to argue with a percentage of 3.6 %. On the other hand, the lowest aspect is the skills to make decisions at 2.7 %.
The percentage of critical thinking skills was still in the very low category because the average achievement for each question with the HOTS indicator was only 24 %. Furthermore, the aspect of decision-making skills was 17 % in the very low category. In the argumentative skills, the indicators consisted of (1) claim, (2) data, and (3) rebuttal. 80 % of students in the high HOTS category correctly answered questions by following the data indicator. In the rebuttal indicator, only 40 % of students answered correctly. Lastly, in the claim indicator, 60 % of students answered correctly.

Achievement of HOTS Aspects for Students with Low HOTS
The numbers of students in the low HOTS category were ten students. The percentage of achievement in this category was 31 %. The highest percentage was 48 % in this category, and the lowest percentage was 26 %. The students' low HOTS was caused by unfamiliarity with working on HOTS questions, as shown by the results of the questionnaire analysis on statement 6, which obtained a percentage of 29 % in the poor category. This result is coherent with (Anggraini et al., 2019;Rahayuningsih & Jayanti, 2019). They state that students could not complete HOTS because of difficulty determining ideas to explain them. This research is also supported by (Heong et al., 2012;Yee et al., 2015) that difficulties in generating ideas experienced by students will cause them to experience technical problems in completing their assignments. Another factor is the unfamiliarity with the HOTS questions. (Hall & Piazza, 2008;Lavy, 2020;Rochman & Hartoyo, 2018) state that other factors influencing students' low HOTS are culture and character. Different cultures and characters will affect the students' mindset.
For students with low HOTS, the order of achievement of the highest HOTS aspects to the lowest is (1) critical thinking skills (7.6 %), (2) problemsolving skills (6.2 %), (3) decisionmaking skills (6.2 %), (4) argumentative skills (5.8 %), and (5) creative thinking skills (5.2 %). Critical thinking skills were the aspect that got the highest achievement, although it was still classified in the very low criteria (25 %). (Febriana & Sinaga, 2021;Samsul Hadi et al., 2018;Priyadi et al., 2018) claim in their research that the students' low critical thinking skills are caused by several things, including (1) students had difficulty in completing and answering the questions given and (2) students had difficulty in connecting the results of calculations with the phenomena presented. This claim is supported by the results of the questionnaire analysis on statement items number 9 and 10 which obtained a percentage of 33 % and 37 % in the poor category.

The Achievement of HOTS Aspects of Students with poor HOTS
The percentage of students in the poor HOTS category was 53 %. This aspect was the most dominant than other HOTS categories. The aspect that obtained the highest percentage of achievement was the critical thinking skills, with the highest percentage of 13.2 % and the lowest percentage of 7 %. The lowest aspect was the creative thinking skills. At a glance, the results for students with poor HOTS were similar to those of students with low HOTS. However, the difference lies in the percentage of decision-making and argumentative skills. For students with poor HOTS, argumentative skills were higher than decision-making skills. On the other hand, students in the low HOTS category had higher decision-making skills than argumentative.
Another difference was the achievement of the creative thinking aspect. The percentage of achievement for students in the very low category was 13 %, while for students in the low HOTS category was 17 %. The achievement percentage in both categories was poor because it was only 25 %. For students in the poor HOTS category, their creative thinking skills were lower than those in the low HOTS category.
Low creative thinking skills can be caused because students are unfamiliar with questions that train creative thinking skills. Besides, the learning methods do not train students' creative thinking skills (Rochman & Hartoyo, 2018). Creative thinking skills can be trained by improving learning strategies and models. This argument is supported by (Mahanal et al., 2019;Rohim & Susanto, 2012;Saputri et al., 2019), who state that the improvement of learning strategies is to change the learning model that can facilitate communication between students and students and teachers and students. Also, to foster students' creative thinking skills.

CONCLUSION
The Testlet model assessment instrument integrated with the E-ujian website is feasible for measuring the junior high school students' HOTS. The developed assessment instrument measures students' HOTS based on several characteristics, namely (1) valid, (2) reliable, (3) discriminatory, (4) disproportionate, (5) relevant, (6) representative, 7) impractical because it has many obstacles when implementing it, and (8) specifically functions to measure students' HOTS with aspects measured in the form of critical thinking skills, creative thinking skills, problem-solving skills, decision-making skills, and argumentative skills; specifically for the object of science.
This research can be used as a reference for teachers or students of the educational program as one of the innovations of learning assessment instruments that can be applied during the Covid-19 pandemic. The development of the Testlet model assessment instrument can also be used to measure the five skills of HOTS: (1) critical thinking skills, (2) creative thinking skills, (3) problemsolving skills, (4) argumentative skills, and (5) decision-making skills.