English or languish - Probing the ramifications
of Hong Kong's language policy
   
 
Multiple Discriminant Analysis (MDA)    
     
statistical modelling / multiple discriminant analysis | 3-step analysis (step1 | step 2)

 

Step 3 - Interpretation

  • Introduction - Interpretaion of the discriminant function involves a close examination of the discriminant weights, the discriminant loadings, and the partial - F values.

  • Discriminant weights (discriminant coefficients) - The magnitude and sign of the discrimant weights provide information about the relative importance and direction of each independent variable on the value of the discriminant function's dependent variable (discriminant score).

    A small weight can mean either that the independent variable is unimportant in determining the discriminant score, or that it's importance has been negated by the presence of other independent variables with which it shares a high degree of correlation.

    Unstable descriminant weights should be treated with caution.

  • Discriminant loadings (structural correlations) - These are simple linear correlations between each of the independent variables and the value of the discriminant function. In general, discriminant loadings are considered more reliable than discriminant weights in measuring the relative importance of each independent variable on the value of discriminant scores.

  • Partial - F values - There are two basic approaches to computing discriminant functions

    • Simultaneous computation, and
    • Stepwise computation

When stepwise computation is employed partial-F statistics are generated. These statistics can be ranked and compared in very much the same way that discriminant weights are ranked and compared. Partial-F statistics are advantageous in so far as they measure both statistical significance and relative importance.

 
top