Specifies the maximum number of PROX iterations to be performed. The PROX, "normal approximation", algorithm provides starting values for the JMLE algorithm, in order to speed up estimation. PROX has little influence on the final JMLE estimates. PROX iterations will always be performed so long as inestimable parameters have been detected in the previous iteration, because inestimable parameters are always dropped before the next iteration. At least 2 PROX iterations will be performed. PROX iteration ceases when the spread of the persons and items no longer increases noticeably (0.5 logits). The spread is the logit distance between the top 5 and the bottom 5 persons or items.

 

If you wish to continue PROX iterations until you intervene with Ctrl+F, set MPROX=0. JMLE iterations will then commence.

 

At least 1 PROX iteration is performed, and perhaps several more when the data are incomplete.

 

Example 1: To set the maximum number of PROX iterations to 20, in order to speed up the final JMLE estimation of a symmetrically-distributed set of parameters,

 MPROX=20

 

Example 2: To minimize the influence of PROX on the final JMLE estimates,

 MPROX=2 ; this is the minimum possible number of PROX iterations.