The measurement of the target variable, here attitude extremity, seems a little shaky to me.
First of all, using min and max values is not very smart because of outliers and complete disregard of the "coherence" of the opinions, i.e., if standard deviation is small or big (comparatively).
Secondly, the formula is not symmetric. Maybe that's ok because the distributions aren't but using the maximum in the denominator of the attitude extremity formula just increases the effect my first point.
I think a definition in terms of percentiles would have been probably much better. This would implicitly define attribute extremity in terms of non-conformity but that's IMO exactly what it is, it would just sound less dangerous and more 'revolutionarish'.
First of all, using min and max values is not very smart because of outliers and complete disregard of the "coherence" of the opinions, i.e., if standard deviation is small or big (comparatively).
Secondly, the formula is not symmetric. Maybe that's ok because the distributions aren't but using the maximum in the denominator of the attitude extremity formula just increases the effect my first point.
I think a definition in terms of percentiles would have been probably much better. This would implicitly define attribute extremity in terms of non-conformity but that's IMO exactly what it is, it would just sound less dangerous and more 'revolutionarish'.