modify xaxis with different ranges (some very close to 1.xx and others from 1.5 to 10)

[antor82]

Hi all 

I attach the code and the resulting figure

proc sgplot data=aha.forest noautolegend nocycleattrs;
  styleattrs datasymbols=(squarefilled);
  scatter y=predictors x=or / xerrorupper=ucl xerrorlower=lcl group=grp;
  yaxistable or lcl ucl p / y=predictors location=inside position=right;
  refline 1 / axis=x noclip;
  xaxis type=log max=10 minor display=(nolabel)  valueattrs=(size=7) valuesformat=f2.0;
  yaxis display=(noticks nolabel)  fitpolicy=none reverse valueshalign=left
        colorbands=even colorbandsattrs=Graphdatadefault(transparency=0.8)  valueattrs=(size=7);
run;

I need to amplify the x axis around 1 (let's say from 1.01 to 1.1) and then restart with a log scale.

Can you help me please?

Tks in advance

Antonio

[quickbluefish]

Are you saying you want the range 1.01 to 1.1 to be on a linear scale?  

[antor82]

Maybe on a linear scale I will need a very large image to account for a clear visibility…. A double scale (linear at the beginning and log thereafter) could be more effective but I've read that this is not possible. I’ve tried a linear scale with broken axes but that gave me a very narrowed first part of the graph without any increase in visibility

[FreelanceReinh]

Hi @antor82,

Would it help to create a broken x-axis? Last week, @BrunoMueller posted a nice solution demonstrating how this can be done. Here's my attempt to adapt his code to your requirements:

/* Store repetitive code elements in macro variables */

%let plot=scatterplot x=or y=predictors / subpixel=off primary=true group=grp
          xerrorupper=ucl xerrorlower=lcl legendlabel="predictors" name="scatter";
%let xopts=type=log griddisplay=on display=(ticks tickvalues) tickvalueattrs=(size=7);
%let yopts=reverse=true labelfitpolicy=split tickvalueattrs=(size=7) tickvaluehalign=left type=discrete 
           discreteopts=(tickvaluefitpolicy=none colorbands=even
                         colorbandsattrs=graphdatadefault(transparency=0.8));

/* Create the GTL template with a manually broken x-axis */

proc template;
  define statgraph forest_break;
    begingraph / datasymbols=(squarefilled);
      /* the column weights determine the space available */
      layout lattice / columns=3 rowdatarange=union columnweights=(0.35 0.25 0.4) ;
        layout overlay / 
          xaxisopts=(&xopts logopts=(viewmin=0.6 viewmax=0.8
                                     minorticks=true tickvaluelist=(0.6 0.7 0.8)))
          yaxisopts=(display=(tickvalues line) &yopts)
          y2axisopts=(labelFitPolicy=Split);
          &plot;
        endlayout;
        layout overlay / 
          xaxisopts=(&xopts logopts=(viewmin=0.9 viewmax=1.11
                                     minorticks=true tickvaluelist=(0.9 0.95 1 1.05 1.1))
                     offsetmin=0.02)
          yaxisopts=(display=none &yopts);
          &plot;
          ReferenceLine x=1 / clip=false;
        endlayout;
        layout overlay / 
          xaxisopts=(&xopts logopts=(viewmin=1.16 viewmax=10
                                     minorticks=true tickvaluelist=(1.2 2 3 4 5 6 7 8 9 10))
                     offsetmin=0.0075)
          yaxisopts=(display=none &yopts);
          &plot;
          innermargin / align=right;
            axistable value=or  y=predictors / labelposition=min display=(label);
            axistable value=lcl y=predictors / labelposition=min display=(label);
            axistable value=ucl y=predictors / labelposition=min display=(label);
            axistable value=p   y=predictors / labelposition=min display=(label);
          endinnermargin;
        endlayout;
      endlayout;
    endgraph;
  end;
run;

/* Use the template to create the graph */

ods graphics / reset=all;

proc sgrender data=aha.forest template=forest_break;
run;

Result (omitting the titles for brevity):

(Note the spelling of the word "length".)

[quickbluefish]

Wow, that's impressive.