Outlier detection based on robust parameter estimates
Outliers can influence the analysis of data in various different ways. The outliers can lead to model misspecification, incorrect analysis results and can make all estimation procedures meaningless. In regression analysis, ordinary least square estimation is most frequently used for estimation o...
保存先:
| 主要な著者: | , , , |
|---|---|
| フォーマット: | 論文 |
| 言語: | English |
| 出版事項: |
2017
|
| 主題: | |
| オンライン･アクセス: | http://eprints.unisza.edu.my/5995/1/FH02-ICODE-18-13380.pdf http://eprints.unisza.edu.my/5995/ |
| タグ: |
タグ追加
タグなし, このレコードへの初めてのタグを付けませんか!
|
| 要約: | Outliers can influence the analysis of data in various different
ways. The outliers can lead to model misspecification, incorrect
analysis results and can make all estimation procedures
meaningless. In regression analysis, ordinary least square
estimation is most frequently used for estimation of the
parameters in the model. Unfortunately, this estimator is
sensitive to outliers. Thus, in this paper we proposed some
statistics for detection of outliers based on robust estimation,
namely least trimmed squares (LTS). A simulation study was
performed to prove that the alternative approach gives a better
results than OLS estimation to identify outliers. |
|---|