Finding Starting Values for the Item Parameters ·and Suitable Discrimination Indices in the One-Parameter Logistic Model

Suitable values for these indices can be estimated from (part) of the data by a weighted least squares algorithm. It turns out that the same algorithm can also be used with a slight modification to obtain good initial estimates of the item parameters. The algorithm is applicable for binary as well as for polytomous data, in complete as well as in incomplete designs. In case of complicated designs a two-step procedure can be used, which needs far less computer storage than the original algorithm. Two examples are given.