The Backfitting Algorithm

1. Initialize:

         b = average(y)
         set f1, f2, .., fp by do regression with y and x1, y and x2, ..., y and xp

2. Iteration:

   f1 = S1( y -b - f2 - f3 -..-fp | x1) :  S1 denotes a smooth of the residuals-1 against  the predictor x1
   f2 = S2( y -b - f1 - f3 -..-fp | x2) :  S2 denotes a smooth of the residuals-2 against  the predictor x2
     ...
   fp = Sp( y -b - f1 - f2 -..-fp-1 | xp) :  Sp denotes a smooth of the residuals-p against  the predictor xp

3. Repeat 2. until the function don't change
           

留言

這個網誌中的熱門文章

標準差與 Wald 統計量

可能性比檢定(Likelihood ratio test)

Wold Decomposition Theorem