Density plot with group option

[marghe]

Hello,

I have been having troubles with density plots in sgplot/sgpanel. My dataset has 4 subgroups and, when I use the density statement with the group option, a density within each group is computed without taking in consideration the proportion of each subgroup with the whole dataset. How can I "resize" these curves, so that each density reflects its actual size with respect to the whole dataset?

Thanks in advance.

[PaigeMiller]

I am not sure I understand the request. But please look at this example and scroll down to the second example, is that what you want? https://documentation.sas.com/?docsetId=grstatproc&docsetTarget=p0o7d7xxvzs9qmn1vctlufpz1448.htm&doc...

[marghe]

Thanks for your reply but unfortunately I couldn’t find what I was looking for. I’ll make an example to better explain :
I need to analyse the distribution of a continuous variable (let’s call it X) according to these 4 subgroups:

TRT GENDER #SUBJECTS
A M 20
A F 30
B M 10
B F 40
____
100
If we take the first subgroup “A - M”, for example, the density of X will be computed on a total of 20 subjects instead of 100. As a result from the procedure, we will get 4 density functions conditional to each subgroup; while I would like to have a joint density function on (X,TRT,GENDER) - made up 4 densities. It’s really up to the total used to do computations.
Hope this example gives you some more insight on my question.
Thank you!

[PaigeMiller]

Well, probably I still don't understand. Can you explain why the example I mentioned is NOT what you want?

[ballardw]

---

@marghe wrote:

Hello,

I have been having troubles with density plots in sgplot/sgpanel. My dataset has 4 subgroups and, when I use the density statement with the group option, a density within each group is computed without taking in consideration the proportion of each subgroup with the whole dataset. How can I "resize" these curves, so that each density reflects its actual size with respect to the whole dataset?

Thanks in advance.

---

Please at least show the code you have attempted. Without that we have no clue what options you may already be using (or not using) that might be part of a solution. Post the code in a code box opened with the forum's </> or "running man" icon to preserve formatting.

And it would help a lot to have a set of data to display.

Instructions here: https://communities.sas.com/t5/SAS-Communities-Library/How-to-create-a-data-step-version-of-your-dat... will show how to turn an existing SAS data set into data step code that can be pasted into a forum code box using the <> icon or attached as text to show exactly what you have and that we can test code against.

Only include the variables needed to make the plot.

[marghe]

Thanks for your replies, here is my code (it's easier for me to use a sashelp dataset):

data dataset;
set sashelp.class;
run;

/*Age categorization*/
data class;
set dataset;
if Age<=13 then Age_cat='<=13';
else Age_cat='>13';
run;

/*Sex and Age category combined to get a group with 4 levels*/
data class_group;
set class;
if Sex='M' then do;
if Age_cat='<=13' then cat='M - <=13';
else cat='M - >13';
end;
else do;
if Age_cat='<=13' then cat='F - <=13';
else cat='F - >13';
end;
run;

proc sort data=class_group;
by cat;
run;

/*Density plot by group (Plot1)*/
title 'Height of boys and girls according to age category';
proc sgplot data=class_group;
density height /group=cat type=kernel;
xaxis label='HEIGHT';
run;

/*From class_group, 5 subjects have cat='F - <=13' and the total dataset has 19 subjects*/
/*In a histogram, frequencies are computed within each subgroup (Plot 2)*/
title 'Height of boys and girls according to age category';
proc sgplot data=class_group(where=(cat='F - <=13'));
histogram height /group=cat binwidth=10;
xaxis label='HEIGHT' values=(50 to 70 by 10);
run;

Plot1

Plot 2

The red lines indicate the bars' heights if frequency of 'F - <=13' was computed on the total dataset (19 subjects) instead of the size of the subgroup (5 subjects).

I would like that in the computation of densities for each group these 19 subjects were used, not only those belonging to that particular subgroup. Do I have to use weights?

Hope this clarifies a bit my question, thanks you all!!