Date:         Fri, 18 Apr 2008 15:34:08 -0400
Reply-To:     "P. Cristian Gugiu" <crisgugiu@YAHOO.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "P. Cristian Gugiu" <crisgugiu@YAHOO.COM>
Subject:      Proc Univarite (scale of graphs)

Hi all. I'm getting addicted to all the learning I getting from this list
group. Okay, I finally collected the data set I was interested in and
decided to look at the histograms of three of my variables of interest: N,
Skewness, Kurtosis. The scale of my graphs is rather ugly because each
variable has a couple of extreme outliers. I have a couple studies whose N
exceeds 1000, so my variable has an extreme right skew. My histogram of
skewness is a little better, but two studies I included had vars with a skew
of 18 and 42. Finally,  the histogram of kurtosis is heavily right-skewed
due to two vars with a kurtosis value of 400 and 3200. One solution may be
to just get rid of these extreme outliers, which I can afford to due because
I collected moments data on 1199 variables. However, I would like to know if
there is a way to control the scale of the histogram, say skewness could
range between +/- 6. Kurtosis could range between -2 and 20. Sample size
could range between 10 and 500.

Also, how can I output the percent of moments that fall within specified
ranges. So, I have skew values for 1199 variables. How many fall within the
range -0.25 and 0.25.

Thanks,
Cristian