Dominance Analysis SAS Macros

References for general information on dominance analysis:

  • Budescu, D. V. (1993), Dominance analysis: a new approach to the problem of relative importance of predictors in multiple regression, Psychological Bulletin, 114, 542-551;
  • Azen, R., and Budescu D. V. (2003). The dominance analysis approach for comparing predictors in multiple regression. Psychological Methods, 8, 129-148.

These macros were written in SAS (version 8).

THE PROGRAMS:

mini-graphics-computer-401093   DA tables macro (Click to download)

  • This program outputs the additional contributions table as discussed in Budescu (1993).
  • The OUTPUT of this program includes the following information:
    • Regression analysis using the input data file;
    • Dominance analysis results for each dominance level:
      • Additional contributions table;
      • Average contributions within model size;
      • Overall averaged contributions.

 

mini-graphics-computer-401093  DA macro (Click to download)

  • This program outputs the additional contributions table as discussed in Budescu (1993) as well as Dij values, means, standard errors and reproducibility results as discussed in Azen & Budescu (2003).
  • The OUTPUT of this program includes the following information:
    • Regression analysis using the input data file;
    • Dominance analysis results for each dominance level:
      • Additional contributions table (Complete dominance);
      • Average contributions within model size (Conditional dominance);
      • Overall averaged contributions (General dominance).
    • For each of the above dominance levels:
      • Frequencies and probabilities of dominance over repeated sampling;
      • Mean and SE of dominance indicator over repeated sampling;
      • Reproducibility of sample results over repeated sampling.

 

mini-graphics-computer-401093  Multivariate DA macro (Click to download).

  • This program outputs the additional contributions table as discussed in Azen & Budescu (2006).
  • This program will compute DA based on
    1. each of q response variables (Y) separately;
    2. a function of the q response variables taken together using multivariate R2; and
    3. a function of the q response variables taken together using multivariate P2.
NOTE: This program might be quite SLOW!

 

mini-graphics-computer-401093  Logistic DA main program | Logistic DA macros (Click to download — both programs needed).

  • This program provides output for the analysis of empirical data sets as discussed in Azen & Traxel (2009).
  • This program will compute DA for logistic regression using one binary response variable (Y) and p predictor variables.