Author Topic: selection of nodes in multidimensional Rasch Model  (Read 607 times)


  • Newbie
  • *
  • Posts: 9
    • View Profile
selection of nodes in multidimensional Rasch Model
« on: May 18, 2015, 11:46:30 AM »
Hi everyone! my name is Alejandro Veas, I´m researcher at the University of Alicante (Spain), at the Department of developmental psychology and didactics.

I am trying to make some Rasch analysis of a school attitudes test, which has 35 items with 7 categories and 5 dimensions. N= 1400 secondary students. I used rating scale model in multidimensional Rasch form.
My question is: Is any criteria to select the number of nodes?. I´ve used 1000, but I don´t know if that´s correct.


Eveline Gebhardt

  • Administrator
  • Full Member
  • *****
  • Posts: 103
    • View Profile
    • Email
Re: selection of nodes in multidimensional Rasch Model
« Reply #1 on: May 19, 2015, 11:22:39 PM »
Hi Alejandro

The default number of nodes per dimension is 15 (for Gauss and Quadrature). It's probably best to start with this and increase the number of nodes by 5 or 10 if the model does not converge. Another reason for increasing the number of nodes could be a large variance in item thresholds and student scores. The default range of nodes (in the estimate command) is from -6 to 6 logits. If your distribution is wider, you can make this range larger (e.g. minnodes=-10, maxnodes=10) and in that case I would increase the number of nodes.

In case of multidimensional models with more than 3 dimensions, you need to use montecarlo estimation. In this case, you give the total number of nodes over the dimensions. Usually start with 1500 and increase is the model does not converge.

I hope this helps. Please let me know if you have any more questions.