Multiple Histograms/Densities on same graph?

[mariosegal]

My pleasure, attached 4 samples from R, 3 versions that could work, I also added a faceted one which was not what I asked for  and that  I CAN do in sgpanel

the code I used is below, I also had to read the data and create a factor, which is not dissimilar to having a sas format - all in one line with minimal options, will get a bit more complicated if I wanted to format axes or add axis labels (in either platform)

  #chart1

ggplot(small, aes(x=sheetz_num,fill=atm_group)) + geom_histogram(alpha=.2, position="identity",binwidth=1)+labs(title="Chart1")

#chart2

ggplot(small, aes(x=sheetz_num,fill=atm_group)) + geom_histogram(alpha=1, position="dodge",binwidth=1)+labs(title="Chart2")

#chart3

ggplot(small, aes(x=sheetz_num,color=atm_group)) + geom_density()+labs(title="Chart3")

#chart4

ggplot(small, aes(x=sheetz_num)) + geom_histogram(alpha=.2, position="identity",binwidth=1)+labs(title="Chart4")+facet_grid(atm_group~.)

I do not believe this is possible with gplot as suggested, unless I transform the data first with Proc freq (basically creating the bins) so I can have 3 variables as in the stock example, but why should I do that, for that matter I can maybe have 2 powerpoint charts on top of each other somehow, it would be cumbersome and time consuming.

I appreciate the help, and I do it because I like SAS and would like to be able to do it in SAS,

[GraphGuy]

Thanks for the examples! 

Here's some code showing how you could create something like Chart1 using gplot.

Yes, you'll have to pre-summarize your data.

One suggestion - once you get a plot you like, you might want to write a macro so that you can generate the plot by passing parameters to the macro (similar to what you're doing in R) rather than repeating the code each time.

data foo;

input atm_group sheetz_num count;

datalines;

1 1 1950

1 2 1500

1 3 900

1 4 400

1 5 300

2 1 6100

2 2 2400

2 3 1600

2 4 1000

2 5 550

;

run;

axis1 label=(angle=90 'count') minor=none offset=(0,0);

axis2 order=(0 to 20 by 5) minor=none offset=(0,0);

symbol1 value=none interpol=needle width=15 color=Af0582f55;

symbol2 value=none interpol=needle width=15 color=A39dee455;

legend1 position=(right middle) across=1 label=(position=top);

title "Chart1";

proc gplot data=foo;

plot count*sheetz_num=atm_group /

vzero vaxis=axis1 haxis=axis2 legend=legend1

autohref autovref chref=grayee cvref=grayee;

run;

[blchri]

I had the same problem as your Chart 3 last year.  One solution (to chart 3) is to create your own template file.  It's kind of like a layout() in R.   This one will create 3 densities overlaid on one plot.  I can give you a full example if you want.

proc template;

   define statgraph d;

   begingraph;

   layout overlay / xaxisopts=(label='Volume');

   densityplot eval((log(src = 'bc') + 1) * vol) /

      kernel() lineattrs=(color=purple);

   densityplot eval((log(src = 'bp') + 1) * vol) /

      kernel() lineattrs=(color=red);

   densityplot eval((log(src = 'fp') + 1) * vol) /

      kernel() lineattrs=(color=gray);

   endlayout;

   endgraph;

   end;

run;

proc sgrender data=abcd template=d;

run;

[mariosegal]

Thank you all - it is clear exactly how to do it in SAS now

What I want to emphasize is that in R I am not doing any pre-processing, nor am I required to create a template or a macro, I am just using ggplot2 - the equivalent in SAS would be to do it in sgplot/sgpanel/gplot without pre-processing. The simple calls I showed do all the work

If I worked at SAS, implementing the grammar of graphics and allowing this sort of things would be on my priority list

I really like SAS, I believe it can do most of what people need, and every day I discover new things (The ability to call support for help is a super valid feature, the R forums are not quite the same). And it deals with large datasets easily while in R we are constrained to in memory or have to use special techniques - but graphics is one area where it lags a bit - as shown here - it is hard to always have labels on top of bars as well, if they do not fit perfectly they just are not printed.

[blchri]

If you feel that R graphics are better for your needs, you can actually call R from SAS using IML studio.   Here's some information about calling R from SAS:

http://support.sas.com/rnd/app/studio/statr.pdf

[mariosegal]

I did not know that, very interesting.

I am not familiar with IML - is that part of Base SAS?

[blchri]

It's a separate product.  I'm not sure about all the details.  I think the name changed in the last release or two.  IMLplus and/or IML Studio depending on the release?  would know.

SAS IML website:

Interactive matrix programming, SAS/IML | SAS

[Rick_SAS]

You can call R from PROC IML and from SAS/IML Studio.

For a video that shows the basics, see Video: Calling R from the SAS/IML Language - The DO Loop

For an example and discussion, see Calling R from SAS/IML software - The DO Loop