Nalazite se na CroRIS probnoj okolini. Ovdje evidentirani podaci neće biti pohranjeni u Informacijskom sustavu znanosti RH. Ako je ovo greška, CroRIS produkcijskoj okolini moguće je pristupi putem poveznice www.croris.hr
izvor podataka: crosbi !

Polychoric correlation coefficient in forecast verification (CROSBI ID 569043)

Prilog sa skupa u zborniku | sažetak izlaganja sa skupa | međunarodna recenzija

Pasarić, Zoran ; Juras, Josip Polychoric correlation coefficient in forecast verification. Helsinki: FMI, 2009

Podaci o odgovornosti

Pasarić, Zoran ; Juras, Josip

engleski

Polychoric correlation coefficient in forecast verification

Forecast verification based on KxK contingency tables is not yet standardized. In the measure-oriented approach various scores are calculated and used to condense some aspects of the forecast quality, each score resulting in a single number. The final goal is to assess particular forecasting system or to compare various such systems. Beside classical measures like the Heidke or the Pierce ones, the Gandin-Murphy family of scores are used. The latter includes the Gerrity and the LEPS sub-families. On the other side, verification problem is multifaceted and no score can comprehend all information that is contained in the contingency table. For this reason the distribution-oriented approach has been proposed by Murphy and Winkler. Here, the joint empirical distribution of forecasts and observations as given by the KxK table is analyzed as a whole. In the present work a measure of association in the KxK table, known in social sciences as polychoric correlation coefficient (PCC) is applied. A standardized bivariate normal distribution is related to the table in a natural way. This normal distribution is fully specified by its correlation coefficient which in turn is the PCC of the table. The PCC possesses several desirable properties including the weak sensitivity on the number of categories. Moreover, from the bivariate normal distribution, which is determined by the PCC, and from the marginal frequencies, it is possible to reconstruct fairly well the original table. In this way the dimensionality of the problem is reduced from KxK to 2K, while the differences between the original and the reconstructed table could be further analyzed from the distributional point of view. The method is systematically applied to a large set of 6x6 contingency tables on verification of quantitative precipitation forecasts.

Contingency tables; tetrachoric correlation coefficinet

nije evidentirano

nije evidentirano

nije evidentirano

nije evidentirano

nije evidentirano

nije evidentirano

Podaci o prilogu

2009.

objavljeno

Podaci o matičnoj publikaciji

Helsinki: FMI

Podaci o skupu

4th International Verification Methods Workshop

predavanje

08.06.2009-10.06.2009

Helsinki, Finska

Povezanost rada

Geologija