Multiple outliers detection: application to research & development spending and productivity growth
Multiple outliers are frequently encountered in applied studies in business and economics. Most of the practitioners depend on ordinary least squares (OLS) method for parameter estimation in regression analysis without identifying outliers properly. It is evident that OLS totally fails even in prese...
| Egile Nagusiak: | , |
|---|---|
| Formatua: | Artikulua |
| Hizkuntza: | English |
| Argitaratua: |
BRAC University
2010
|
| Gaiak: | |
| Sarrera elektronikoa: | http://hdl.handle.net/10361/421 |
| Gaia: | Multiple outliers are frequently encountered in applied studies in business and economics. Most of the practitioners depend on ordinary least squares (OLS) method for parameter estimation in regression analysis without identifying outliers properly. It is evident that OLS totally fails even in presence of single outlying observation. Single observation outlier detection methods are failed to numerically compare the sensitivity of the most popular diagnostic statistics. Data set from Griliches and Lichtenberg (1984) is used to show that we need to take extra care for model building process in presence of multiple outliers. |
|---|