NotesFAQContact Us
Collection
Advanced
Search Tips
ERIC Number: ED317601
Record Type: Non-Journal
Publication Date: 1990-Apr
Pages: 25
Abstractor: N/A
Reference Count: N/A
ISBN: N/A
ISSN: N/A
Feel No Guilt! Your Statistics Are Probably Robust.
Micceri, Theodore
This paper reports an attempt to identify appropriate and robust location estimators for situations that tend to occur among various types of empirical data. Emphasizing robustness across broad unidentifiable ranges of contamination, an attempt was made to replicate, on a somewhat smaller scale, the definitive Princeton Robustness Study of 1972 to determine how closely results produced in a laboratory environment represent the multiple contaminations encountered among real world data. Contaminations included various mixtures of modalities, digit preferences, tail-weights, sample spaces, and asymmetry. Due at least partly to the almost universal presence of asymmetry, the arithmetic mean in particular and L-estimators in general proved comparatively robust for the situations investigated. Most so-called "robust" estimators proved less efficient than the mean even in rather extreme conditions for these multinomial data sets produced by empirical applications of ability and psychometric measures. These findings imply that prior robustness studies that have found the arithmetic mean and its parametric counterparts to be non-robust may be misleading, since the types of theoretical populations investigated in most research studies do not appear to exist among real world psychometric and education data sets. Seven data tables and two graphs are included. (Author/TJH)
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A