Purwanto, Adhitya Dimas Eko (2019) Studi Simulasi : Perbandingan Metode Imputasi Mean Substitution Dan Median Substitution Pada Person Trait Estimation. Sarjana thesis, Universitas Brawijaya.
Abstract
Missing data adalah istilah yang digunakan ketika terdapat informasi yang hilang dalam penelitian. Missing data dapat diatasi menggunakan teknik imputasi dengan cara menggantikan data yang hilang dengan nilai yang diperkirakan cukup layak. Penelitian ini bertujuan untuk mengetahui perbandingan metode mean substitution dan median substitution untuk mengatasi missing data dalam mengestimasi person-trait. Penelitian ini merupakan penelitian studi simulasi dengan menggunakan pendekatan Item Response Theory (IRT) dengan model penskoran Graded Response Model (GRM) serta ilustrasi pada real data. Studi simulasi dilakukan dengan dua kondisi jumlah item yang berbeda yaitu 20 item dan 40 item dengan format respons polytomous. Setiap kondisi item yang berbeda tersebut akan dikondisikan dengan persentase missing data sebesar 10%, 20% dan 50% dengan jumlah responden sebanyak 1000 responden. Kemudian estimasi person-trait dilakukan dengan expected a posteriori (EAP) berdasarkan model GRM dan analisis RMSE. Hasil estimasi person-trait akan dikorelasikan menggunakan pearson product moment. Hasil penelitian secara keseluruhan menunjukan bahwa metode median substitution lebih efektif digunakan apabila persentase missing data 10% dan 20%. Sedangkan metode mean substitution efektif digunakan pada persentase missing data 50%. Selain itu, keefektifan penggunaan metode imputasi dipengaruhi oleh jumlah item yang digunakan dan persentase missing data
English Abstract
Missing data is the term used when there is information which missing in a test. Missing data can be solved using an imputation by replacing missing with an estimated value. This study aims to determine the effectiveness of the mean substitution and median substitution method to overcome missing data in estimating person-trait. This research is a simulation study using the Item Response Theory (IRT) approach with a Graded Response Model (GRM) scoring model and an illustration on real data. Simulation studies are carried out with two conditions, the number of different items, namely 20 items and 40 items in the format of the polytomous response. Each condition of the different items will be conditioned with 10%, 20% and 50% of missing data with the number of respondents as many as 1000 respondents. The person-trait estimation is done with the expected a posteriori (EAP) based on the GRM model and RMSE analysis. The person-trait estimation results will be correlated using pearson product moment. The overall result of the study showed that median substitution method is more effective when the percentage of data missing is 10% and 20%. While the mean substitution method is more effective in the percentage of missing data by 50% In addition, the effectiveness of using the imputation method is influenced by the number of items used and the percentage of missing data.
Other obstract
-
Item Type: | Thesis (Sarjana) |
---|---|
Identification Number: | SKR/FISIP/2019/794/051908213 |
Uncontrolled Keywords: | Graded Response Model, Item Response Theory, Missing Data, Imputation, Mean substitution, Median substitution |
Subjects: | 100 Philosophy, parapsychology and occultism, psychology > 150 Psychology > 150.2 Miscellany > 150.287 Testing and measurement |
Divisions: | Fakultas Ilmu Sosial dan Ilmu Politik > Psikologi |
Depositing User: | Budi Wahyono Wahyono |
Date Deposited: | 17 Nov 2020 17:46 |
Last Modified: | 19 May 2022 04:15 |
URI: | http://repository.ub.ac.id/id/eprint/174783 |
Text
Adhitya Dimas Eko Purwanto.pdf Restricted to Registered users only Download (2MB) |
Actions (login required)
View Item |