A paper of Ingels and Maenhout (2016), which is currently under revision, discusses the impact of proactively scheduled overtime and reactively allocated unscheduled overtime on the robustness of a personnel shift roster. In this respect, the trade-off between the workforce size and scheduled overtime and between scheduled overtime and unscheduled overtime is considered. 

The dataset underlying the computational experiments in this paper consists of 27 requirements classes, which correspond to 27 different (TCC, DCD, SCD)-combinations (Vanhoucke and Maenhout 2009). Each class consists of 10 files, which define the expected minimum staffing requirements (R_{dh}^w) and the extra staffing requirements (R_{dh}^{w,extra}) for 7 days and 6 demand periods. R_{dh}^w and R_{dh}^{w,extra} respectively define the minimum and extra staffing requirements during demand period h on day d. Each file consists of two 7 x 6 matrices that respectively define R_{dh}^w and R_{dh}^{w,extra} with the rows indicating the days and the columns indicating the demand periods.