Anderson, C.J, Verkuilen, J. & Peyton, B.L. (2009, inpress). Modeling polytomous responses using simultaneously estimated multinomial logistic regressions. Journal of Educational and Behavioral Statistics
Request a copy of this paper.

---

Code is available for:

SAS | LEM | R graphics |

---

SAS:

• Multinomial logistic regression for item A.
• Models reported in Table 1 and some extras.
• Models reported in Table 3 and some extras.

LEM:

• Models reported in Table 1:
  • Independence log-linear model.
  • All 2-way interaction (homogeneous association) log-linear model.
  • One dimensional LMA.
  • Two dimensional LMA.
  • Two dimensional hybrid LMA.
  • Two dimensional hybrid LMA with nu111 = nu121 and nu221=nu231.
  • Two dimensional hybrid LMA with nu111 = nu121.
  • Bock's Nominal response model (MMLE).
  
  
• Models reported in Table 3:
  • Independence log-linear model.
  • All 2-way interaction (homogeneous association) log-linear model.
  • One dimensional LMA + lambda*ijt.
  • Two dimensional LMA + lambda*ijt.
  • Hybrid LMA + lambda*ijt.
  • Hybrid LMA + lambda*ijt with nu111 = nu121.

R scripts & data for graphics:

• Multinomial model for item A:
  • R script for plot.
  • data used by script.
• Fitted probabilities of item responses:
  • R script for control group.
  • R script for treatment group.
  • Fitted probabilities for control group (data).
  • Fitted probabilities for treatment group (data).
  • Values of theta used for x-axis (data).
  
  

  ---

  
  

  Report broken links and send email to cja@illinois.edu
  
  

  Last revised May 28, 2009