Solved
Contributor
Posts: 46

# Histogram standart desviation

Hi all, I've been trying to make a graph using PROC UNIVARIATE, however I couldn't understand why is not possible  show de first, second and the third standart desviation as the image below:

How can I show the S.D like the image above, using proc UNIVARIATE ?

Accepted Solutions
Solution
‎06-06-2013 11:44 AM
SAS Employee
Posts: 8

## Re: Histogram standart desviation

Starting with SAS 9.3m2, PROC UNIVARIATE enables you to place reference lines at the values of selected statistics, including multiples of the sample standard deviation.  Here is some code and output.  I have specified a list of bin midpoints so all the reference lines are displayed.  (Axis ranges are not extended automatically to include reference lines.)

data foo;
do i = 1 to 100;
x = 10 + rannor(123);
output;
end;
run;

proc univariate data=foo noprint;
histogram x /
normal(noprint)
statref= -3std -2std -1std mean 1std 2std 3std
statreflabel= "-3 S.D" "-2 S.D" "-1 S.D" "" "1 S.D" "2 S.D" "3 S.D"
midpoints= 7 to 13 by 1
nocurvelegend;
run;

All Replies
SAS Super FREQ
Posts: 1,210

## Re: Histogram standart desviation

This is a custom graph, that you can make with SAS 9.3.  Use PROC MEANS to get all the statistics you need and merge with original data used for histogram.

You can use SGPLOT procedure to create the histogram, and use REFLINE statement to overlay the lines for the SD values.  then, you can annotate the rest of the formation.

Contributor
Posts: 46

## Re: Histogram standart desviation

Hi Sanjay, could you send me a example ?

Doesn't SAS have a function to know the "z-scores" ?

Solution
‎06-06-2013 11:44 AM
SAS Employee
Posts: 8

## Re: Histogram standart desviation

Starting with SAS 9.3m2, PROC UNIVARIATE enables you to place reference lines at the values of selected statistics, including multiples of the sample standard deviation.  Here is some code and output.  I have specified a list of bin midpoints so all the reference lines are displayed.  (Axis ranges are not extended automatically to include reference lines.)

data foo;
do i = 1 to 100;
x = 10 + rannor(123);
output;
end;
run;

proc univariate data=foo noprint;
histogram x /
normal(noprint)
statref= -3std -2std -1std mean 1std 2std 3std
statreflabel= "-3 S.D" "-2 S.D" "-1 S.D" "" "1 S.D" "2 S.D" "3 S.D"
midpoints= 7 to 13 by 1
nocurvelegend;
run;

Contributor
Posts: 46