yang posted on Tuesday, January 07, 2014 - 10:29 am
I'm using Mplus version 7 to estimate unidimensional graded response models with ML. I learned from the manual that by setting "integration = standard(48)", Mplus will use rectangular quadratures (am I correct?). My question is how Mplus determines the range of quadrature nodes?
With rectangular integration (our default) we use equally spaced integration points between –5 to 5
With Gauss Hermite integration the integration points are roots of the Hermite polynomials – they are not equally spaced so technically talking about range is misleading. For 30 points the smallest/largest roots are +-9.7
Also because we use adaptive integration, the integration changes from one iteration to the next and from one observation to another.