A heuristic approach to link some unlinkable tests

The ability to place different tests on a common metric, that is, test equating, is essential for many purposes in psychological and educational research. When traditional linking designs are not viable, however, the ability to perform equating is compromised. To solve this problem, authors have proposed the use of alternative equating procedures based on the identification of pseudocommom items or collateral data. Here we propose a heuristic allowing for the estimation of the intercept linking parameter using as the sole input the Rasch item difficulties of the tests. Using simulated data, the heuristic showed high accuracy and efficiency in estimating the intercept linking parameter (MAE = 0.15 logit; RMSE = 0.21 logit; R2= .87). Test length and true-parameter size had a small-to-moderate impact on the accuracy of parameter estimation. Results are discussed in light of implications for practical purposes, as well as future improvements of the proposed heuristic.