Ranking of part worths and average importances differs

Hi,

I have carried out a CBC and am now investigating it. I have carried out a HB using the Lightstudio analysis tool and calculated the part worth values and average importances of the characteristics.

I noticed that feature A has a part worth of 25.93 with an avg. importance of 12.42%, while feature B has a part worth of 26.65 with an avg. importance of 12.34%. So overall the "ranking" of my characteristics differs in these two positions, depending which metric I choose.

My question is how does this fit together?
Part worth: A (25.93) < B (26.65)
but
Avg. Importance: A (12.42%) > B (12.34%).

Where does this come from and how should this be interpreted?

Many thanks
Tom

Resolved
2 replies