The Sensitivity of the Number of Correctly Forecasted Events to the Threat Score:A Practical ApplicationSource: Weather and Forecasting:;2004:;volume( 019 ):;issue: 003::page 646Author:Hilliker, Joby L.
DOI: 10.1175/1520-0434(2004)019<0646:TSOTNO>2.0.CO;2Publisher: American Meteorological Society
Abstract: The sensitivity (S) of the number of correctly forecast events (?hits?) to changes in threat score is explored. An emphasis is placed on the practical utility of S for end users of operational systems who may need to further assess a system's performance whenever its threat score is provided. It is shown that S is a function of three parameters: (i) the actual threat score, (ii) forecast bias, and (iii) number of events in the sample size. All are easily obtainable if the forecasting system's 2 ? 2 contingency table is also provided. It is also shown that the greatest change in S occurs for threat scores that are relatively low (i.e., close to 0). Conversely, a forecast system that performs well (i.e., has a high threat score) has the least amount of change in S. As a result, only a nominal increase in the number of hits is realized for some incremental increase in threat score. Thus, it is difficult to show continuous improvements for forecast products that perform well already. Examples demonstrating the practical utility of S are presented in this study.
|
Collections
Show full item record
| contributor author | Hilliker, Joby L. | |
| date accessioned | 2017-06-09T15:06:09Z | |
| date available | 2017-06-09T15:06:09Z | |
| date copyright | 2004/06/01 | |
| date issued | 2004 | |
| identifier issn | 0882-8156 | |
| identifier other | ams-3439.pdf | |
| identifier uri | http://onlinelibrary.yabesh.ir/handle/yetl/4172167 | |
| description abstract | The sensitivity (S) of the number of correctly forecast events (?hits?) to changes in threat score is explored. An emphasis is placed on the practical utility of S for end users of operational systems who may need to further assess a system's performance whenever its threat score is provided. It is shown that S is a function of three parameters: (i) the actual threat score, (ii) forecast bias, and (iii) number of events in the sample size. All are easily obtainable if the forecasting system's 2 ? 2 contingency table is also provided. It is also shown that the greatest change in S occurs for threat scores that are relatively low (i.e., close to 0). Conversely, a forecast system that performs well (i.e., has a high threat score) has the least amount of change in S. As a result, only a nominal increase in the number of hits is realized for some incremental increase in threat score. Thus, it is difficult to show continuous improvements for forecast products that perform well already. Examples demonstrating the practical utility of S are presented in this study. | |
| publisher | American Meteorological Society | |
| title | The Sensitivity of the Number of Correctly Forecasted Events to the Threat Score:A Practical Application | |
| type | Journal Paper | |
| journal volume | 19 | |
| journal issue | 3 | |
| journal title | Weather and Forecasting | |
| identifier doi | 10.1175/1520-0434(2004)019<0646:TSOTNO>2.0.CO;2 | |
| journal fristpage | 646 | |
| journal lastpage | 650 | |
| tree | Weather and Forecasting:;2004:;volume( 019 ):;issue: 003 | |
| contenttype | Fulltext |