Deteksi Kesalahan Besar Dengan Kombinasi Tes Statistik Dan Evaluasi Besaran Redundansi
ABSTRAK Outlier detection based on residual analysis by tau test does not imply the actual situation. Sometimes, the actual number of measurements suffered from outliers is less than the detected number. In this research, tau test is conducted, along with evaluation of redundancy value, to figure ou...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Article NonPeerReviewed |
| 出版: |
[Yogyakarta] : Pusat Antar Universitas (PAU) Studi Ekonomi UGM
2004
|
| 主題: | |
| 在線閱讀: | https://repository.ugm.ac.id/21575/ http://i-lib.ugm.ac.id/jurnal/download.php?dataId=4441 |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | ABSTRAK
Outlier detection based on residual analysis by tau test does not imply the actual situation. Sometimes, the actual number of measurements suffered from outliers is less than the detected number. In this research, tau test is conducted, along with evaluation of redundancy value, to figure out the mutual influences between measurements. Evaluation is performed by determining redundancy values of measurements having been detected to suffer from outliers.Tthe investigation is conducted by : (1) simulating one outlier in the network and (2) simulating two outliers in the network.
The results show that both in the network having one oulier and two outliers, tau test detect more outliers than the actual one. Based on the redundancy analysis, it is found that those measurements have high mutual influences, se that outlier detection by tau test only is not enough. By combining with redundancy analysis and tau values, the number and location of outliers can be determined more accurately.
Keyworks: Tes statistik, redudansi, kesalahan besar |
|---|