Controversy in Determining Criteria and Categories in Summarizing and Exploring The Research Data; Analysis of Assessment Procedures in The Social Science Research

Determination of criteria and categorization of data in measurement is a crucial moment when processing research data. Overview, exploration and presentation of data will be effective and valid if the criteria set meets the requirements, both from a theoretical and statistical perspective. However, the fact that often arises, especially in beginner researchers made a mistaken in determining the criteria and categories. These errors are often indicated in the form of incomprehension and negligence of researchers in placing data types with the type of methodology. This paper describes various errors in determining criteria and categories of research data, determining criteria and categories of research data, and suggesting their meaning, especially in social science research. This research involved respondents from four different types of assessments, which is 428 respondents from Daharnis-Zadrian Majoring Choices Inventory (DZ-MCI) and 174 respondents from the Acceptability of Mental-Health MobileApp Survey (AMMS), as well as 1190 respondents from the development of DLAS and DLESS inventory. The results of the study show that the determination of categories and criteria in data processing greatly impacts the truth of data exposure and advanced data analysis. Implications of these studies for determining categories and criteria of research on social science research are discussed.


Introduction
Many assessors and researchers, especially the amateur ones have difficulty in determining the criteria or categorization of research results that have been obtained (Meyer, Doromal, Wei, & Zhu, 2017). Sometimes they use criteria from the results of previous studies that are considered "true" even though the categorization is actually "wrong". There are also those who adopt raw assessment criteria for an educational institution that are not in accordance with the objectives of the research conducted ( This article aims to provide an overview of how criteria are determined from research based on criterion-referenced measurement and normreferenced measurement that meets scientific and logical requirements, also based on statistical analysis.

Participants
The sample in this study was based on the distribution of respondents according to the type PSYCHOLOGY AND EDUCATION (2021)

Data Analysis
Research data analysis was carried out in accordance with various alternative categorization and grouping of data based on comparison with ideal (criterion-referenced measurement) and inter-respondent scores (norm-referenced measurement). Data grouping based on score distribution is done by utilizing Microsoft Excel software (Triola, 2013) and Statistical Package for Social Science (SPSS). Data analysis was also carried out using Rasch analysis (Sumintono & Widhiarso, 2014) through Winstep Version 3.72 software (Linacre, 2009).

Result
Data grouping and categorization can be done in two times, which are before the process of collecting research data using ideal score to establish criterion-referenced measurement and after data is collected by using real data through norm-referenced measurement ( In various sources of data analysis instructions that are popular with academics and researchers, categorization and grouping of data can be done by searching for class intervals and class limits (Lipsey et al., 2017;Shapiro et al., 2017). These results will be made justification about the position of the respondent on the variable being measured, for example in a study, researchers will measure the condition of student achievement motivation in a school, the researcher can categorize the position of respondents which to be known as "high", "moderate", " low" ( For example, the application of criteria for the procedure for 5 scaling instruments (1-5 scale), and the establishment of five categories can be seen in the results of research on the conditions of student learning activities in West Sumatra-Indonesia whose the data was obtained from the Daharnis Learning Activities Scale (DLAS) with 30 items as shown in Table 1. In the grouping process the lowest and highest score (ideal) becomes the benchmark in determining the data class, where 30 is the lowest score and 150 is the highest score on a 1-5 scale. For data sourced from instruments other than 1 to 5 and categorization other than five the category can adjust according to the designation procedure. Based on the steps to determine the category, the group of data can be presented as follows: However, the determination of the steps in Table 1 still reaps quite different controversies and interpretations among academics and researchers in interpreting respondents' achievements, especially in respondents who achieved a threshold score between two categories. This condition is often discussed in the term "statistical casualties" in the data.
The illustration of the case is if a respondent gets a score of 101 on a DLAS measuring instrument, then he will be included in the "moderate" category, whereas if there are other respondents who get a score of 102 then they will get a "high" category. In this case, the score achievement is only 1 point adrift between the two respondents, so that this condition tends to be detrimental and wrong in interpreting a data. Therefore, related to processing and interpretation using these categories, there is a need for in-depth analysis supported by other data to see the condition of the respondents in a holistic and detailed manner. Alternative of data grouping and categorization can also be done by using real / empirical data for establishing norm-referenced measurement. One application that can be used is the Rasch Model (Hagquist, Bruce, & Gustavsson, 2009; Seo, Taherbhai, & Paek, 2013). In the Rasch Model, real/empirical data is assumed to be ordinal data that is transformed into equal interval data. The transformation was carried out through calculating opportunities in each choice of answers in the form of odd ratio data in the form of opportunities that can be obtained in each answer option calculated based on logarithms. The result of the transformation is in the form of an index of separation amount that will determine how many ideal categories.
The illustrations of determining data categories based on Rasch analysis are presented in Figure 1, whose data is obtained from the Acceptability of Mental-Health Mobile App Survey (AMMS) administration. In Figure 1 it can be seen that the data separation index is 3.20. The numbers of groups in the AMMS data category are 3 categories, namely "high", "medium", and "low" groups. Through the index separation, the researcher can group data using the mean and standard deviation.
The categorization using this approach can be seen in Table 2. The explanation in Table 2 can be interpreted that the grouping of data carried out by referring to the norm standard is not too much different than the benchmark on the ideal score. The illustration of the grouping category is presented in Figure 2.

MEASURE
Person -MAP -Item  Figure 2 The example of categories from Rasch perspective The exposure in Figure 2 shows the division of categories based on the standard deviation value. The data distribution is displayed in the form of a normal curve in the person graph. This curve is then divided at point +1SD (+1.61) to -1SD (-1.61) so that respondents whose logit values are in this range are categorized as respondents with "moderate" achievement. Transforming this value to raw score can also be seen in the table, that in the category "moderate" is the respondent with raw score 22.2 to 34. The condition of the data with this norm reference will also change according to the conditions of empirical data and the quality of the measuring instrument itself. Therefore, it is important for researchers to determine what approach will be used in determining the process of grouping and categorizing research data

Conclusion
In determining the category of data distribution can be used two standards, namely criterionreferenced measurement and norm-referenced measurement. The selection of these types of benchmarks is based on the theoretical basis and research objectives. If the researcher wants to compare the position of the respondent with ideal conditions, then criterion-referenced measurement is used. Meanwhile, if researchers want to compare positions between respondents, normreferenced measurement is used. For the purposes of evaluation, intervention and treatment, it will be better if criterion-referenced measurement is used. However, if for the sake of selection it is used as a standard norm-referenced measurement.