This web page provides SAS programs and other information for the paper: 

Bilder, C. R. and Loughin, T. M. (2007).  Modeling Association between Two or More Categorical Variables that Allow for Multiple Category Choices.  Communications in Statistics: Theory and Methods 36(2), 433-451. 

If you find the programs below useful, please let me know in an e-mail to chris@chrisbilder.com!  I would be interested in hearing about your use of the programs and the MRCV data that you are modeling. 

• Additional simulation results
• Non-convergence example for maximum likelihood estimation

• These files correspond to the marginal modeling approach examples given in Section 4.1 of the paper.
  • control_2MRCV_final.sas - This is the control program for the statistical methods.  Use this program to run the programs boot_2MRCV_final.sas, data_gen_resample.sas, and gange_2MRCV.sas. 
  • 2_MRCV_example_data.txt - Data set for test waste and waste storage only.
  • boot_2MRCV_final.sas - Fits a generalized loglinear model via the marginal modeling method and use the bootstrap and Rao-Scott approximations to find a p-value for testing the model of interest vs. the saturated model. 
  • gange_2mrcv.sas - Finds a vector of multinomial probabilities subject to the constraints of the H_o model. 
  • data_gen_resample.sas - Using a set of multinomial probabilities from Gange.sas, this program generates the resamples under the H_o model. 
• These files correspond to the marginal modeling approach examples given in Section 4.2 of the paper.
  • control_3MRCV_final.sas - This is the control program for the statistical methods.  Use this program to run the programs boot_3MRCV_final.sas, data_gen_resample.sas, and gange_3MRCV.sas. 
  • 3_MRCV_example_data.txt - Data set for test waste, waste storage, and sources of veterinary information MRCVs.
  • v_mat_m.csv - File containing the V matrix (see Appendix A) used in the PROC IML code of boot_3MRCV_final.sas and read in control_3MRCV_final.sas (see program comments for more information). 
  • boot_3MRCV_final.sas - Fits a generalized loglinear model via the marginal modeling method and use the bootstrap and Rao-Scott approximations to find a p-value for testing the model of interest vs. the saturated model.
  • gange_3MRCV.sas - Finds a vector of multinomial probabilities subject to the constraints of the H_o model. 
  • data_gen_resample.sas - Using a set of multinomial probabilities from Gange.sas, this program generates the resamples under the H_o model.  This is the same program as listed in the Section 4.1 programs.   

This work is supported in part by National Science Foundation grants SES-0207212 and SES-0233321.