Help on linear-plateau regression model

[sysmars]

we want to use this model to calculate the X threshold where Y begin to decline, with conidence interval.

The attached is out data.

who has the procedure? could ou send me a copy? thanks a lot.

[SteveDenham]

Carefully look at the documentation for PROC NLIN, especially Example 63.1 Segmented model.  This fits a quadratic up to an unknown join point, followed by a plateau.

title 'Quadratic Model with Plateau';

proc nlin data=a;

  parms alpha=.45 beta=.05 gamma=-.0025;

  x0 = -.5*beta / gamma;

  if (x < x0) then

  mean = alpha + beta*x + gamma*x*x;

  else mean = alpha + beta*x0 + gamma*x0*x0;

  model y = mean;

  if _obs_=1 and _iter_ =. then do;

  plateau =alpha + beta*x0 + gamma*x0*x0;

  put / x0= plateau= ;

  end;

  output out=b predicted=yp;

run;

I hope this gets you started on what you wish to do.

Steve Denham

[sysmars]

Thanks so much Steve, and also, does below procedure can also be used for our data? what the difference between this 2 procedures?

DATA a;

INPUT met gain @@;

DATALINES;

80 102 85 115 90 125 95 133 100 140

105 141 110 142 115 140 120 142

;

PROC NLIN;

PARMS a = 102 b = 2.7 c = -0.04;

x = met-80;

x0 = -.5*b / c;

IF x < x0 THEN

MODEL gain = a+b*x+c*x*x;

ELSE

MODEL gain = a+b*x0+c*x0*x0;

IF _obs_=1 and _iter_ =. THEN DO;

plateau = a+b*x0+c*x0*x0;

x0 = x0+80;

PUT / x0 = plateau= ;

END;

RUN;

and another question is, how to make a graph like below, thanks a lot.