I want to run a discriminant analysis on different motion capture measures to see which of the measures distinguishes best between my two conditions. The problem is that some of the standardized discriminant function coefficients are >1 and <-1. How is that possible? What did I do wrong? I used z-transformed data to make sure the scales are no issue. Thanks for your help.
Asked
Active
Viewed 414 times
1
-
standardization, in general, does not guarantee the quantity being standardized will lie between -1 and 1. – Macro Jan 23 '12 at 19:19
-
To have unstandardized discriminant coefficients equal the standardized ones one must scale the predictor variables to variance 1 within groups (did you do that?). But the standardized coefficients can exceed 1 in magnitude. Do not confuse them with regression coefficients beta. – ttnphns Jan 23 '12 at 19:39