The LORELIA residual test: a new outlier identification test for method comparison studies
In this work, a new outlier identification test for method comparison studies based on robust linear regression is proposed in order to overcome the special problem of heteroscedastic residual variances. Method comparison studies are performed in order to prove equivalence or to detect systematic di...
Saved in:
| Main Author: | |
|---|---|
| Format: | Book/Monograph Thesis |
| Language: | English |
| Published: |
2009
|
| Subjects: | |
| Online Access: | Verlag, kostenfrei, Volltext: http://elib.suub.uni-bremen.de/diss/docs/00011612.pdf Langzeitarchivierung Nationalbibliothek, Volltext: http://d-nb.info/998406295/34 Resolving-System, Volltext: http://nbn-resolving.de/urn:nbn:de:gbv:46-diss000116129 
|
| Author Notes: | by Geraldine Rauch |
| Summary: | In this work, a new outlier identification test for method comparison studies based on robust linear regression is proposed in order to overcome the special problem of heteroscedastic residual variances. Method comparison studies are performed in order to prove equivalence or to detect systematic differences between two measurement methods, instruments or diagnostic tests. They are often evaluated by linear regression methods. As the existence of outliers within the dataset can bias non robust regression estimators, robust linear regression methods should be preferred. In this work, the use of Passing-Bablok regression is suggested which is described in [Passing, Bablok, 1983], [Passing, Bablok, 1984] and [Bablok et al. 1988]. Passing-Bablok regression is a very outlier resistant procedure which takes random errors in both variables into account. Moreover, the measurement error variances are not required to be constant, so Passing-Bablok regression is still appropriate if the error variances depend on the true concentration which is a common situation for many laboratory datasets ... |
|---|---|
| Physical Description: | Online Resource |