Dear Philip

Thank you for your comments. After some hunting, I stumbled on the option that allows one to appoint a group variable. For those others who may not be aware of this,

1) right click on the group column 2) you will be presented with several options. Select "group" 3) Select the variables to be plotted 4) select scatter plot under analyze

Insight appears to be somewhat inconsistant in offering options since the aforementioned method is offered but some analysis tools but in at least one other case, that of XY fit, a window appears that allows one to make several choices about the analysis including selecting a group variable.

Thanks again for the tip.

Nat

Philip Whittall <philip.whittall@UNI To: SAS-L@LISTSERV.UGA.EDU LEVER.COM> cc: Sent by: "SAS(r) Subject: Re: Proc Output: A Proposal for a new procedure Discussion" <SAS-L@LISTSERV.UGA. EDU>

02/11/02 12:53 PM Please respond to Philip Whittall

Dear Nathaniel,

I clearly did not understand this one. When I checked scatter plots in INSIGHT (version 8.2) there is a GROUP box in the options and if you put a varable in there you get a scatter plot for each distinct of that variable.

Philip

PS the new robust/resitant regression routines in IML are excellent for this type of screening as they can detect multiple outliers.

On Mon, 11 Feb 2002 09:03:26 -0500, Nathaniel Wooding <Nathaniel_Wooding@DOM.COM> wrote:

>Last week I encountered a situation wherein I wanted to split a SAS data >set into a number of smaller sets, each according to a single value of an >id variable. The goal was to then use the scatter option of SAS/Insight to >examine each of these datasets for outliers and identify the >characteristics of the outliers. > >For those who wonder why I might need to do this, I had fish lengths and >weights for a half dozen or so species and I was looking for keypunch >errors. The screening is best done for each individual species since >differences in body characteristics give different length-weight >relationships. The scatter option of Insight does not offer a means (or I >can't find it) of producing separate plots for each value of a group >variable -- hence, the need for multiple datasets. One could, of course, >use the 'hide in plots' feature of the SAS/Insight spreadsheet but that >would be a bit of a pain in this circumstance. > >It was simple enough to code a data step which split the file into the >necessary pieces and I may write a macro to do this automatically. However, >it would seem useful for SAS to offer a Proc which would do this operation. >It might need to allow one to provide a prefix for the data set names as >Proc Transpose does for new column titles. > >Does anyone else have a need for such a Proc or have any comments. I did >look at SQL to see if it could split a data set in this manner but I did >not find a solution. > >______________________________________________________ >Nat Wooding >Environmental Specialist III >Dominion, EP&C >4111 Castlewood Rd >Richmond, VA 23234 >Phone:804-271-5313, Fax: 804-271-2977