For an equal-precise measure, items must be selected so that the test information curve is relatively flat across the trait range. This means that a set of highly discriminating items with a broad range of difficulty parameters must be identified. If there are certain trait ranges where the test information is not high enough, new items will need to be written specifically for that purpose, and these new items will again have to be administered to a heterogeneous calibration sample. How much information is high enough? Well, if the conditional information is around 10, then the conditional standard error is about 0.31. To place this in more conventional terms, a standard error of 0.31 corresponds to a reliability coefficient of 0.90.