Hi Can someone please explain why is it that the higher the Gini coefficient, the more predictive the model is? My understanding is that in order for Gini coefficient to be = 1, the model curve (according to the gains curve shown in the notes) needs to be convex towards the top right hand side, so that the: area between the model curve and the diagonal / area of the triangle above the diagonal = 1. Am I right?
Yes, you're right. I thought this website http://www2.cs.uregina.ca/~hamilton/courses/831/notes/lift_chart/lift_chart.html gave quite a good explanation of why a big area gives a more 'predictive' model. Also, though, bear this thread in mind: http://www.acted.co.uk/forums/showthread.php?t=3988. To be honest, it's such a small part of the Core Reading that I would hope that simply an awareness of the curve would be sufficient. It's certainly never been examined in ST8 (so far!).