Date:         Sat, 25 Dec 2010 08:02:27 -0500
Reply-To:     Peter Flom <peterflomconsulting@mindspring.com>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Peter Flom <peterflomconsulting@MINDSPRING.COM>
Subject:      Re: cluster bar chart with +/- 1 standard error
Comments: To: Ali Zanaty <zanaty2005@YAHOO.COM>
Content-Type: text/plain; charset=UTF-8

Ali Zanaty <zanaty2005@YAHOO.COM> wrote


<<<
I am a new SAS user and currently, I am teaching an undergraduate course in statistics.

If possible, I need some help with

(1) Create a cluster bar chart

(1) add plus/minus 1 standard deviation to each bar in cluster bar chart
both sides and upper side only.

(2) add * over each pair of bars to indicate the averages are statistically differ (p value<0.05)
>>>

Nat Wooding already gave you a start on this.

But bar charts are not good graphics, and adding error bars makes them worse, and clustering doesn't help either.  These are known as "dynamite plots" because 1) they look like the devices that detonate dynamite and 2) They are dangerous and can blow up.

See:
biostat.mc.vanderbilt.edu/twiki/pub/Main/TatsukiRcode/Poster3.pdf

and

http://pablomarin-garcia.blogspot.com/2010/02/why-dynamite-plots-are-bad.html

for example.

Dot plots are better if you have relatively few data, and box plots or one of their variants if you have lots of data.


Both of these are available in the SGPLOT procedure.  But if you want to trellis your data, you may want to look into SGPANEL.

HTH

Peter

Peter L. Flom, PhD
Statistical Consultant
Website: http://www DOT statisticalanalysisconsulting DOT com/
Writing; http://www.associatedcontent.com/user/582880/peter_flom.html
Twitter:   @peterflom