You may save the fitting value within a precudure, then graph it in poc gplot.

Here is an example,

  title 'U.S. Population Growth';
   data uspop;
      input pop :6.3 @@;
      retain year 1780;
      year   = year+10;
      yearsq = year*year;
      datalines;
   3929 5308 7239 9638 12866 17069 23191 31443 39818 50155
   62947 75994 91972 105710 122775 131669 151325 179323 203211
   226542 248710
   ;


   title 'Beaton/Tukey Biweight Robust Regression using IRLS';
   proc nlin data=uspop nohalve;
      parms b0=20450.43 b1=-22.7806 b2=.0063456;
      model pop=b0+b1*year+b2*year*year;
      resid = pop-model.pop;
      sigma = 2;
      k     = 4.685;
      if abs(resid/sigma)<=k then _weight_=(1-(resid / (sigma*k))**2)**2;
      else _weight_=0;
      output out=c r=rbi p=pop_pred;
   run;

   proc print; run;

   symbol1 i=join v=square color=green ;
   symbol2 i=join v=circle color=red;


   proc gplot data=c;
   plot (pop pop_pred)*year/overlay legend;
   run;
   quit;

不是很清楚啊

谢谢。我也学到了一招

3# bobguy

受教了。。

2# bobguy 非常感谢！