Date:         Thu, 14 Aug 2003 13:17:32 -0400
Reply-To:     Jay Weedon <jweedon@EARTHLINK.NET>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Jay Weedon <jweedon@EARTHLINK.NET>
Organization: http://extra.newsguy.com
Subject:      Re: Computing points for plots in regression
Content-Type: text/plain; charset=us-ascii

On Thu, 14 Aug 2003 08:41:42 +0000 (UTC), wajnberg@antibes.inra.fr
(Eric Wajnberg) wrote:

>
>Dear all,
>
>I am using PROC REG to compute a quadratic regression. ok.
>
>I am now willing to make a graph ploting simultaneously: (1) the data points,
>(2) a curve showing the fit (predicted values) and (2) two lines (one above
>and the other below the fit) showing +-SE of the fit.
>
>In a certain way, what I need is what SAS uses in the symbol statment with
>I=RQCLM... (but I do not whant confidence intervals, I only want SE).
>
>I need to collect all the corresponding values to make the graph using another
>software more efficient that SAS/Graph.
>
>How I can do that?
>
>Using only PROC REG gives me the values I want, but only at the observed data,
>and not all along all potential values of the independant variable.
>
>On the other hand, PROC SCORE give me only the predicted value (for another
>data set) but not the SE.
>
>Can somone help me on that?
>
>Thanks in advance for any help on that.

proc reg data=mydata;
  model y=x xsquared;
  output out=temp p=pred stdp=stderr;
run; quit;

data temp;
  set temp;
    upper=pred+stderr; lower=pred-stderr;
run;

proc gplot data=temp;
  symbol1 i=none c=black v='*'; symbol2 i=rq c=red v=none;
  symbol3 i=rq c=blue v=none; symbol4 i=rq c=green v=none;
  plot (y pred upper lower)*b /overlay;
run;

JW