topic Re: Proc gmap questions in Graphics Programming

Proc gmap questions

sasgorilla — Tue, 05 Aug 2025 00:55:27 GMT

Hi. First time using proc gmap to build a world map to visualize number of studies on a topic.

Here's where I am at with a trial image. It appears to be doing what I believe it should do thus far.

I do have a couple warnings in my log I wanted to ask about.

4267  title;
4268  /* Step 5: Generate choropleth map */
4269  proc gmap map=merged_map_clean data=merged_map_clean all;
4270      id idname;
4271      choro study_count2 / levels=5
4272                          coutline=grayaa  /* gray borders for countries WITH studies */
4273                          missing
4274                          legend=legend1;
4275      format study_count2 studyfmt.;
4276     run;

WARNING: Some observations were discarded when charting study_count2.  Only first matching
         observation was used.  Use STATISTIC= option for summary statistics.
WARNING: Some observations were discarded when charting study_count2.  Only first matching
         observation was used.  Use STATISTIC= option for summary statistics.
NOTE: Foreground color WHITE same as background. Part of your graph might not be visible.
NOTE: 47729 bytes written to C:\Users\XXXXXXXXXXXXX\gmap30.png.
WARNING: The specified value of 16.5833 inches for HSIZE= is larger than 16.0000 inches which
         is the maximum for the device WIN. HSIZE is ignored.
WARNING: The specified value of 8.7188 inches for VSIZE= is larger than 8.3958 inches which is
         the maximum for the device WIN. VSIZE is ignored.
4277  quit;

1. Would the first set of warnings refer to the multiple rows for each country from the mapsgfk.world map I merged with? That code is here:

proc sort data=mapsgfk.world out=map_sorted; by idname; run;

data merged_map_clean;
    merge map_sorted(in=a) country_fixed(in=b);
    by idname;
    if a;  /* Keep all countries from map */
    
    /* Exclude Antarctica */
    if upcase(idname) = "ANTARCTICA" then delete;

    /* Create categorized study count variable */
    if missing(study_count) then study_count2 = 0;
    else if study_count = 1 then study_count2 = 1;
    else if 1 < study_count <= 5 then study_count2 = 2;
    else if 5 < study_count <= 10 then study_count2 = 3;
    else if 10 < study_count then study_count2 = 4;
run;

2. Where are the size warnings coming from? I haven't set a size anywhere and can't figure how, or if I need to change.

Also, I've attached the countries and study counts I used to populate my image in this test run.

Thanks for your help!

Re: Proc gmap questions

whymath — Tue, 05 Aug 2025 02:26:16 GMT

Anwser to question 1:

The warning message comes from the statistic= option:

STATISTIC=FIRST | SUM | FREQUENCY | MEAN

specifies the statistic for GMAP to chart. For character variables, FREQUENCY is the only allowed value--any other value is changed to FREQUENCY and a warning is issued. The frequency of a variable does not include missing values unless the MISSING option is specified.

FIRST
GMAP matches the first observation from the DATA= data set and charts the response value from this observation only. This is the default. If more rows exist that are not processed, a warning is issued to the log.

SUM
All observations matching a given ID value are added together and the summed value is charted.

FREQUENCY
A count of all rows with nonmissing values is charted unless you specify the MISSING option.

MEAN
All observations matching a given ID value are added together and then divided by the number of non-missing observations matched. This value is then charted unless you specify the MISSING option.

By adding statistic=first to choro statement, the warning message is gone.

Re: Proc gmap questions

Ksharp — Tue, 05 Aug 2025 03:58:01 GMT

You want this ?

data world(index=(idx=(id Segment)));
set mapsgfk.world(where=(CONT ne 97)); *Get rid of Antarctica Continent;
if ID='US' then group=4;  *USA;
 else if ID='CA' then group=3;  *Canada;
  else if ID='BR' then group=2;  *Brazil;
   else if ID='AU' then group=1;  *Australia;
     else if ID=:'CN' then group=1;  *China;
	  else if ID='GB' then group=3; *United Kindom;
	   else if ID='DE' then group=4; *Germany;
	    else group=0; *Others;
run;
data world2;
   set world;
   by id Segment;
   if first.Segment then PolyID+1;    /* create ID variable for polygons */
run;
proc sort data=world2;
by group;
run;
proc format;
value fmt
0='0 studies'
1='1 studies'
2='2-5 studies'
3='6-10 studies'
4='>10 studies'
;
run;

ods graphics/width=1000px height=600px;
proc sgplot data=world2 noborder /*aspect=0.5 */ ;
 styleattrs datacolors=(white CXBDD7E7 CX6BAED6 CX3182BD CX08519C);
   polygon x=x y=y ID=PolyID / fill outline group=group lineattrs=(color=grayaa)  ;
   xaxis display=none offsetmin=0 offsetmax=0;
   yaxis display=none offsetmin=0 offsetmax=0 ;
   keylegend / position=bottom location=outside noborder title='Study Count' ;
   format group fmt.;
run;

Re: Proc gmap questions

sasgorilla — Thu, 07 Aug 2025 22:03:13 GMT

Thank you for the recommendation, @Ksharp .

I lost some of my country data when I tried with your code for some reason.

Re: Proc gmap questions

GraphGuy — Mon, 20 Apr 2026 15:18:41 GMT

Typically, you would not want to merge your data with the map dataset.

I would recommend merging it with mapsgfk.world_attr (which contains 1 observation per country) instead.

FIRST	GMAP matches the first observation from the DATA= data set and charts the response value from this observation only. This is the default. If more rows exist that are not processed, a warning is issued to the log.
SUM	All observations matching a given ID value are added together and the summed value is charted.
FREQUENCY	A count of all rows with nonmissing values is charted unless you specify the MISSING option.
MEAN	All observations matching a given ID value are added together and then divided by the number of non-missing observations matched. This value is then charted unless you specify the MISSING option.