topic Re: Efficiently graph multiple variables (columns) stratified by dichotomous variable in Graphics Programming

Efficiently graph multiple variables (columns) stratified by dichotomous variable

amarikow57 — Wed, 17 Feb 2021 18:24:16 GMT

Is there an efficient way to create either boxplots or histograms for multiple continuous variables at once? Such as a do loop or actual function? I tried to look up a loop for graphing online, but didn't find anything that translated for me.

I do not want the graphs to overlap as is happening with this code:

PROC SGPLOT data = CLEANED.Analytic_Data;
 vbox ADOS2_3_CODINGE_ANXTY_C / category = ATEAM_AUTISM_YES_NO;
 vbox ADOS2_3_SCORESUMM_COMPSCORE / category = ATEAM_AUTISM_YES_NO;
 vbox ADOS2_3_SCORESUMM_RRBTOTAL / category = ATEAM_AUTISM_YES_NO;
 vbox ADOS2_3_SCORESUMM_SATOTAL / category = ATEAM_AUTISM_YES_NO;
 vbox ADOS2_3_SCORESUMM_OVERALL / category = ATEAM_AUTISM_YES_NO;
RUN;

I have about 30 more variables to graph on top of this. I'd rather not have to type out PROC SGPLOT repeatedly if possible. Furthermore, is there a way to get a select group of variables to be side-by-side in a plot (as aforementioned, the above code just overlaps the plots). That is, all variables that start with ADOS2_3 share the same axis in one plot while stratifying the ATEAM_AUTISM_YES_NO, etc.

Thank you in advance.

Re: Efficiently graph multiple variables (columns) stratified by dichotomous variable

Reeza — Wed, 17 Feb 2021 18:32:42 GMT

Reconfigure your data so you can use a BY statement instead to get individual plots?

Flip it using PROC TRANSPOSE so you have the variable names in one column and the values in a second. Then use a single SGPLOT. You can then create some custom versions using a WHERE statement.

proc sgplot data=....;
*optional where;
where variableName like 'ADOS2_3_%';
by variableName;
vbox variableValue / category = ateam_autistm_yes_no;
run;

@amarikow57 wrote:

Is there an efficient way to create either boxplots or histograms for multiple continuous variables at once? Such as a do loop or actual function? I tried to look up a loop for graphing online, but didn't find anything that translated for me.

I do not want the graphs to overlap as is happening with this code:
PROC SGPLOT data = CLEANED.Analytic_Data;
 vbox ADOS2_3_CODINGE_ANXTY_C / category = ATEAM_AUTISM_YES_NO;
 vbox ADOS2_3_SCORESUMM_COMPSCORE / category = ATEAM_AUTISM_YES_NO;
 vbox ADOS2_3_SCORESUMM_RRBTOTAL / category = ATEAM_AUTISM_YES_NO;
 vbox ADOS2_3_SCORESUMM_SATOTAL / category = ATEAM_AUTISM_YES_NO;
 vbox ADOS2_3_SCORESUMM_OVERALL / category = ATEAM_AUTISM_YES_NO;
RUN;
I have about 30 more variables to graph on top of this. I'd rather not have to type out PROC SGPLOT repeatedly if possible. Furthermore, is there a way to get a select group of variables to be side-by-side in a plot (as aforementioned, the above code just overlaps the plots). That is, all variables that start with ADOS2_3 share the same axis in one plot while stratifying the ATEAM_AUTISM_YES_NO, etc.

Thank you in advance.

Re: Efficiently graph multiple variables (columns) stratified by dichotomous variable

WarrenKuhfeld — Wed, 17 Feb 2021 18:30:34 GMT

The macro language and a macro %DO loop will do what you want. https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.5&docsetId=mcrolref&docsetTarget=p0ri72c3ud2fdtn1qzs2q9vvdiwk.htm&locale=en

Re: Efficiently graph multiple variables (columns) stratified by dichotomous variable

djrisks — Wed, 17 Feb 2021 18:36:32 GMT

As Reeza said, you should reconfigure your data, so that you can use the BY Statement. That would stop you have to write so many VBOX statements.

If you want the graphs side by side (in different cells), then you should use SGPANEL, instead of SGPLOT, after you have reconfigured your data.

If you want to also express groups differently on the same cell, then you could look into GROUPDISPLAY = CLUSTER option.

Many thanks,

Kriss

Re: Efficiently graph multiple variables (columns) stratified by dichotomous variable

ghosh — Sat, 20 Feb 2021 23:06:46 GMT

Hard to give an answer without any sample data.

I would transpose the data into three columns (eg. ADOS2_3_type, ADOS2_3_val ATEAM_AUTISM_YES_NO) and use sgpanel to display them side by side:

proc sgpanel data=mydat;
panelby ADOS2_3_type / rows=6 columns=5 ;
vbox ADOS2_3_val / category=ATEAM_AUTISM_YES_NO;
run;