Abstract: The PCM2 sampler has been implemented as a FORTRAN subroutine which can be called by an R function. A demonstration will be given, not only to show the easiness in the use of the method but also to demontstrate some yet unsolved problems. The sampler in its simple form does not sample uniformly from the sampling space. To obtain a uniform distribution (all matrices having the same probability) the basic algorithm is enhanced with the Hastings-Metropolis algorithm which ensures a uniform distribution but at the same time can slow down the walk through the sampling space, by staying in the same state during two or more steps. This slowing down phenomenon is very small in the Rasch
sampler but it is substantial in the PCM2 sampler. It will be discussed how different variations of the basic algorithm can possibly alleviate this problem.
14 Feb. 2013 → 15 Feb. 2013
International Workshop on Psychometric Computing
Österreichische Systematik der Wissenschaftszweige (ÖFOS)