smooth function of wind direction & wind speed on BC concentration

[imsenny]

I want to make a graph showing the effect of both wind speed and wind direction on black carbon concentration - so basically x axis = wind speed y axis = wind direction, and the more red the higher BC concentration etc...

I think I can use proc gam or proc loess.. but not to sure how to code it

Thanks in advance!

[Rick_SAS]

In PROC SPLOT you can just use a LOESS statement or a PBSPLINE statement to overlay a nonparametric curve on a scatter plot.  (You can also use the REG statement for a parametric curve.)

For example:

proc sgplot data=sashelp.enso;
pbspline x=year y=pressure;           /* also draws scatter plot */
loess x=year y=pressure / nomarkers;  /* uses PRESEARCH option */
run;

For more control, you can use SAS procedures. See the article "How to automatically select a smooth curve for a scatter plot in SAS."

[Jay54]

When the real data contains geographical elements like direction, or the data is cyclical, a polar graph can be very useful.

http://blogs.sas.com/content/graphicallyspeaking/2012/04/09/simpler-is-better/

[PGStats]

Ideally you would want a radar chart where angle represents wind direction, radius represents wind speed, and color represents geometric mean black carbon concentration.  That would be nice!

Anybody up to the challenge?

[imsenny]

Yes that's exactly what I want actually.. Something like polarplot in R...

[Jay54]

If you can include some sample data with the appropriate columns and a picture of what you are looking for, you may elicit some suggestions.

[imsenny]

Columns would include - BC concentration, wind speed, and wind direction OR BC concentration, wind direction, and altitude etc.

So different colors will show BC conc, and x-axis wind direction, and y-axis wind speed...

Can SAS do this?

[Jay54 · Accepted solution]

What you are showing is a simple rectangular heat map by speed and direction.  This can be done using SAS 9.40M3 SGPLOT HEATMAP or HEATMAPPARM.  I used SASHELP.HEART to do an example, changed the axes.  It would be more intuitive if the direction was circular.

%let gpath='.';
%let dpi=150;
ods html close;
ods listing image_dpi=&dpi gpath=&gpath;

/*--Replicate data > 1Million obs--*/
data heart;
  set sashelp.heart(keep=systolic diastolic cholesterol);
  output;
  do i=1 to 200;
    systolic=systolic*(1+0.05*(ranuni(2)-0.5));
    diastolic=diastolic*(1+0.05*(ranuni(2)-0.5));
    cholesterol=cholesterol*(1+0.05*(ranuni(2)-0.5));
    output;
  end;
run;

title 'BC Concentration by Wind Speed and Direction';
proc sgplot data=heart(where=(systolic < 250 and diastolic < 150));
  heatmap x=systolic y=diastolic / colorresponse=cholesterol 
          colorstat=mean colormodel=(white green yellow red)
          nxbins=100 nybins=60 name='a';
  gradlegend 'a' / title=''; 
  xaxis  display=(novalues noticks) min=100 max=200 label='Wind Direction';
  yaxis  display=(novalues noticks) min=50 max=125 label='Wind Speed';
run;

[imsenny]

thank you! I haven't tried it out with my data, but think it will work-

[Jay54]

Here is a more intuitive visual for directional data.

http://blogs.sas.com/content/graphicallyspeaking/2016/07/02/polar-graph/