Computerized adaptive tests (CATs) are individualized tests that, from a measurement point of view, are optimal for each individual, possibly under some practical conditions. In the present study, it is shown that maximum information item selection in CATs using an item bank that is calibrated with the one-or the two-parameter logistic model results in each individual answering about 50% of the items correctly. Two item selection procedures giving easier (or more difficult) tests for students are presented and evaluated. Item selection on probability points of items yields good results only with the one-parameter logistic model and not with the two-parameter logistic model. An alternative selection procedure, based on maximum information at a shifted ability level, gives satisfactory results with both models.

Automated test assembly (ATA) has been an area of prolific psychometric research. Although ATA methodology is well developed for unidimensional models, its application alongside cognitive diagnosis models (CDMs) is a burgeoning topic. Two suggested procedures for combining ATA and CDMs are to maximize the cognitive diagnostic index and to use a genetic algorithm. Each of these procedures has a disadvantage: The cognitive diagnostic index cannot control attribute-level information and the genetic algorithm is computationally intensive. The goal of this article is to solve both problems by using binary programming, together with the item discrimination indexes of Henson et al., for performing ATA with CDMs. The three procedures are compared in simulation. Advantages and disadvantages of each are discussed.

Several techniques exist to automatically put together a test meeting a number of specifications. In an item bank, the items are stored with their characteristics. A test is constructed by selecting a set of items that fulfills the specifications set by the test assembler. Test assembly problems are often formulated in terms of a model consisting of restrictions and an objective to be maximized or minimized. A problem arises when it is impossible to construct a test from the item pool that meets all specifications, that is, when the model is not feasible. Several methods exist to handle these infeasibility problems. In this article, test assembly models resulting from two practical testing programs were reconstructed to be infeasible. These models were analyzed using methods that forced a solution (Goal Programming, Multiple‐Goal Programming, Greedy Heuristic), that analyzed the causes (Relaxed and Ordered Deletion Algorithm (RODA), Integer Randomized Deletion Algorithm (IRDA), Set Covering (SC), and Item Sampling), or that analyzed the causes and used this information to force a solution (Irreducible Infeasible Set‐Solver). Specialized methods such as the IRDA and the Irreducible Infeasible Set‐Solver performed best. Recommendations about the use of different methods are given.

Calibration of an item bank for computer adaptive testing requires substantial resources. In this study, we investigated whether the efficiency of calibration under the Rasch model could be enhanced by improving the match between item difficulty and student ability. We introduced targeted multistage calibration designs, a design type that considers ability-related background variables and performance for assigning students to suitable items. Furthermore, we investigated whether uncertainty about item difficulty could impair the assembling of efficient designs. The results indicated that targeted multistage calibration designs were more efficient than ordinary targeted designs under optimal conditions. Limited knowledge about item difficulty reduced the efficiency of one of the two investigated targeted multistage calibration designs, whereas targeted designs were more robust.

In order to optimize measurement precision in computerized adaptive testing (CAT), items are often selected based on the amount of information they provide about a candidate. The amount of information is calculated using item-and person parameters that have been estimated. Usually, uncertainty in these estimates is not taken into account in the item selection process. Maximizing Fisher information, for example, tends to favor items with positive estimation errors in the discrimination parameter and negative estimation errors in the guessing parameter. This is also referred to as capitalization on chance in adaptive testing. Not taking the uncertainty into account might be a serious threat to both the validity and viability of computerized adaptive testing. Previous research on linear test forms showed quite an effect on the precision of the resulting ability estimates. In this chapter, robust test assembly is presented as an alternative method that accounts for uncertainty in the item parameters in CAT assembly. In a simulation study, the effects of robust test assembly are shown. The impact turned out to be smaller than expected. Some theoretical considerations are shared. Finally, the implications are discussed.