Incorporating multi-scenario underreporting rates into MICE for underreported maritime accident record analysis.

This study attempts to incorporate multi-scenario underreporting rates into multivariate imputation using chain equation (MICE) to address underreporting issues in maritime accident records. In addition, the Tobit regression model is utilized to evaluate the economic loss of maritime accidents for both the original and supplemented data. With ten years' maritime accident data in Fujian waters, one empirical study is created to testify the effectiveness of the proposed method. The marginal effect results from the proposed method significantly differ from the conventional Tobit regression model. The proposed method could significantly correct the over- or underestimated impacts related to various variables (e.g., serious/non-serious accident types, fishing vessels, other ships, and liquid cargo ships). This reveals that the proposed method is capable of correcting the systematic deviation of the analysis results to a satisfactory extent. • This study presents a reliable solution to deal with underreporting issues in maritime accident records. • The underreporting issues are transformed into missing data problem through multi-scenario underreporting rates. • The underreporting issues may cause biased maritime accident analysis results. • The systematic deviations of the model results are effectively corrected by the proposed MSUR-MICE method. [ABSTRACT FROM AUTHOR]