1

Recently, I read a paper (Pell et al. 2009) in which the authors use Discriminant Analysis, and I quote:

The discriminant analysis produced three significant canonical functions (Function 1, F(18, 2130) = 164.81, p<0.0001; Function 2, F(10, 1508) = 85.43, p<0.0001; Function 3, F(4, 755) = 22.93, p<0.0001).

I didn't find anything associated with F values of canonical functions in the result of DA done by SPSS, and I cannot understand the F values of the canonical functions. Why are they reported, how are they computed?

amoeba
  • 93,463
  • 28
  • 275
  • 317
Ping Tang
  • 425
  • 1
  • 5
  • 7
  • 1
    SPSS and many other programs of discriminant analysis and canonical correlations (they are closely related) test canonical functions [by Wilk's lambda](http://stats.stackexchange.com/a/74516/3277) which has approximately chi-square distribution. But the latter [is closely related to](http://www.stata.com/support/faqs/statistics/chi-squared-and-f-distributions/) and can be recomptuted onto F distribution. So it is possible to use F. – ttnphns Sep 30 '14 at 12:55

0 Answers0