Assessing Imputation of Extreme Data on Climatological Time Series

Jairo Cugliari (1) , José G Gómez (2)

1 ERIC - Entrepôts, Représentation et Ingénierie des Connaissances
2 LMNO - Laboratoire de Mathématiques Nicolas Oresme

Techniques for imputation of missing data are generally well known. However, they are often applied and validated on non-extreme data. Even for specific approaches specialized on extreme data, the question raises on how to assess the imputation. Because of this, we propose to use known estimators of extreme value theory, such as the ex-tremogram, which we generalize to time series with the presence of missing data.

Domaines: Méthodologie [stat.ME]

Jairo Cugliari, José G Gómez. Assessing Imputation of Extreme Data on Climatological Time Series. Conference on non-stationarity, Jun 2018, Cergy-Pontoise, France. . ⟨hal-01812715⟩