程序
proc nlin data=CCHH_PM method=marquardt;
parms P=0.5 to 1.0 by .02
      k=0.1 to 1.5 by .02;
bounds 0.5<=P<=1.0,0.1<=k<=1.5;
model y=(P*AER*x)/(AER+k);
run;

结果

Note:

An intercept was not specified for this model.

SourceDFSum of SquaresMean SquareF ValueApprox
Pr > F
Model11427194142719423395.1<.0001
Error67140933.661.0039  
Uncorrected Total6721468128   

ParameterEstimateApprox
Std ErrorApproximate 95% Confidence
LimitsLabel
P0.64620.004220.63790.6545
k0.100000.10000.1000
Bound217884.82542.212903.722865.90.1 <= k

感觉很明显对于k值没有换估计的值，一直都是用0.10来拟合，如何克服这个缺点 让程序尽可能多用P K的不同组合来得到并显示SS，因为我的这个模型不可能得到独立的参数估计，只能得到一系列比较好的组合，然后根据理论情况选择