… including normal and global attribution, mult-period multi-attribute etc. etc. I think I have cracked it. It works for me. Involves drawing a picture as follows: 1. Rp and Rb on the y-axis and Wp and Wb on the x-axis. Rp > Rb and Wp > Wb 2. Now join the points such that you have 2 rectangles joined at the origin (R=0, w=0) 3. Now you can do any decomposition based on the areas of the rectangles in your picture. I know it’s hard to visualise and figure out what I am talking about by reading this but if you tell me a site where I can upload a picture, you will know what I am talking about
photobucket.com I’m intrigued about this, I remember you had a good system for simplifying a tricky caluclation in L2.
someone from AF recently posted a link to their blog which explained attribution in a way that sounds similar to your approach. take a look… http://www.houseofgjertsen.info/2011/04/global-portfolio-return-attribution/
see the gjertsen s presentations,google house of gjertsen, I found them useful