Detecting multiple outliers with an application to R&D productivity

Multiple outlying observations are frequently encountered in empirical analyses of economic data. Several multiple outlier tests exist, but little evidence is available on their relative power against alternative causes of outliers and influence. This paper analytically and numerically compares the...

Full description

Saved in:
Bibliographic Details
Published inJournal of econometrics Vol. 43; no. 3; pp. 293 - 315
Main Author Reiss, Peter C.
Format Journal Article
LanguageEnglish
Published Amsterdam Elsevier B.V 01.03.1990
Elsevier
Elsevier Sequoia S.A
SeriesJournal of Econometrics
Subjects
Applications
Econometrics
Economic models
Economic theory
Effects
Exact sciences and technology
Expenditures
Mathematics
Probability and statistics
Productivity
R&D
Regression analysis
Research & development
Research & development expenditures
Sciences and techniques of general use
Statistics
Studies
Variables
Outlier
Productivity
Data analysis
Multiple
Test statistic
Detection
Economic data
Comparative study
Research and development
Online AccessGet full text

Cover

More Information
Summary:Multiple outlying observations are frequently encountered in empirical analyses of economic data. Several multiple outlier tests exist, but little evidence is available on their relative power against alternative causes of outliers and influence. This paper analytically and numerically compares the sensitivity of several outlier diagnostics to different forms of data contamination; it also proposes new statistics. Practical issues associated with these tests are addressed using data on R&D spending and total factor productivity from Griliches and Lichtenberg (1984). The tests confirm their hypothesis that previously low estimates of the return to R&D maybedue to outliers.
ISSN:0304-4076
1872-6895
DOI:10.1016/0304-4076(90)90122-A