Date:         Wed, 3 Jan 2001 21:28:25 GMT
Reply-To:     John Uebersax <jsuebersax@MY-DEJA.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         John Uebersax <jsuebersax@MY-DEJA.COM>
Organization: Deja.com
Subject:      Re: adding std. error bars to plot?

Thanks John (Hixon).

The info you posted is informative, but there's a twist.  I only want
to show the mean value (and standard error bar) at each week, but not
the values for individual patients.  If I understand it, your example
below would plot a point for each case.  Do you know of a way to plot
only, say, a circle for each mean, and a standard error bar through the
circle at each week?

Incidentally, I have also calculated the means and standard errors in a
separate proc, so perhaps these can be used to provide std error bars
in some alternative approach.

John Uebersax


In article <852569C9.006FBCA6.00@knotes.kodak.com>,
  125241N@KNOTES.KODAK.COM wrote:
> From: John Hixon
>
> John asked:
>
> >Date:    Wed, 3 Jan 2001 19:52:31 GMT
> >From:    John Uebersax <jsuebersax@MY-DEJA.COM>
> >Subject: adding std. error bars to plot?
>
> >I would like to add vertical bars reflecting the standard error of
the
> >mean to a plot for the mean value of X over time.  Can someone
suggest
> >how to do this within the context of proc gplot?  thanks in advance.
>
> >John Uebersax
>
> Use the interpolate option for the symbol.  Look up the syntax
> for all kinds of options.  The code below generates an error bar
> that is 2*StdError of the Mean.  You could use 1 SE or 3 SEs etc.
>
> Have fun.
>
> * generate some phony data with 5 groups;
> data sample;
>     do group=1 to 5;
>         do num_samples=ceil(500+100*ranuni(-1));
>             do i=1 to num_samples;
>                 response=100+20*rannor(-1);
>                 output;
>             end;
>         end;
>     end;
> run;
>
> * make the plot and automatically add std error bars;
>
> symbol1 c=blue v=none i=std2tj;
> proc gplot data=sample;
>     plot response*group ;
> run;
>
> I hope this is helpful.  The i=option can do some wonderful things
> on its own, but...nothing replaces a complete analysis of the data
> using Proc Mixed, GLM, or whathaveyou...  :-)
>
> Best regards,
>
> John Hixon
> Eastman Kodak Company
> Rochester, NY USA
>


Sent via Deja.com
http://www.deja.com/