Six computational methods for assembling tests from a bank with an item-set structure, based on mixed-integer programming, are presented. The methods were evaluated using mathematical programming problem feasibility and expected solution times. Two item banks with a set structure from the Law School Admission Test were used to illustrate application of the methods. Solutions were found that satisfied all constraints within short computing times. The best approximations to the target information functions were obtained for the methods that simultaneously selected the stimuli and items for the set-based items.