SAS Macro Programs: agreeplot

The agreeplot macro ( get agreeplot.sas)

Macro program for agreement plots

The AGREEPLOT macro produces a graphic display of agreement in a square n x n contingency table, typically arising when observations are classified on the same scale by two raters or instruments. The display shows both the observed agreement and the expected agreement by superposed shaded and white rectangles, respectively.

This macro provides a more convenient macro interface to the SAS/IML modules for agreement plots, so that you can input a SAS data set (in frequency form), rather than constructing the arrays used by the IML modules. It also provides the means to produce multiple agreement plots for two-way tables stratified by other BY= variables.

In SAS 9.3, these plots are now provided directly by PROC FREQ under ODS graphics, when the AGREE option is specified on the TABLES statement. Other PLOT()=AGREEPLOT sub-options allow some control of the details of these plots.

Parameters

DATA=

The name of the input data set [Default: DATA=_LAST_]

VAR=

Names of factor variables, including BY= variables. The first two variables are used as the rows and columns of the agreement plot.

BY=

Specifies the names of one (or more) By variables. Partial agreement plots are produced for each combination of the levels of the BY= variables. The BY= variable(s) *must* be listed among the VAR= variables.

COUNT=

Name of the frequency variable [Default: COUNT=COUNT]

WEIGHTS=

Specifies weights for exact agreement and for steps removed from exact agreement, both in the plot and in the calculation of the Bangdiwala agreement strength statistic. WEIGHTS=1 counts only exact agreement. WEIGHTS=LIN2 or WEIGHTS=CA2 uses inverse linear (Cicchetti-Allison) weights: 1, 1/(n-1) for an n x n table. WEIGHTS=SQ2 or WEIGHTS=FC2 uses inverse square (Fleiss-Cohen) weights, 1, 1/(n-1)**2. You can also use an IML expression that evaluates to a row vector starting with 1, e.g., WEIGHTS=1 || 1-1/(n-1) or WEIGHTS={1 0.5}.

COLORS=

Specifies colors for exact agreement and for one or more steps removed from exact agreement. [Default: COLORS=BLACK GRAY]

FILLS=

Specifies fill patterns for exact agreement and for one or more steps removed from exact agreement. [Default: FILLS=SOLID LR]

HTEXT=

Height of text value labels. Variable names are scaled up from this by a factor of 1.4 [Default: HTEXT=1.5]

FONT=

Font for text labels. Internal code uses the hardware font HWPSl009 by default for PostScript devices and SWISS otherwise.

LABROT=

Rotation angles for the horizontal and vertical value labels [Default: LABROT=0 90]

TITLE=

Macro program for agreement plots for plot(s). When there are BY= variables, a string indicating their levels is appended.

GOUT=

The name of the graphics catalog [Default: GOUT=WORK.GSEG]

NAME=

The name of the graph in the graphic catalog [Default: NAME=AGREE]

Usage

The AGREEPLOT macro is defined with keyword parameters. The arguments may be listed within parentheses in any order, separated by commas. For example:

  %agreeplot(data=sexfun, var=Husband Wife);

Bugs

With BY= variables, the graphs are correct, but the BN statistics dont use the given WEIGHTS.

Examples

 proc format;
  value rating 1='Never_fun' 2='Fairly_often' 
               3='Very_often' 4='Almost_always';
 data sexfun;
   label Husband = 'Husband rating'
          Wife    = 'Wife Rating';
   format Husband Wife rating.;
   do Husband = 1 to 4;
   do Wife    = 1 to 4;
     input count @@;
     output;
     end; end;
 datalines;
  7     7     2      3
  2     8     3      7
  1     5     4      9
  2     8     9     14
 ;

 *-- Convert numbers to formatted values;
 %table(data=sexfun, var=Husband Wife, char=true, weight=count, out=table);
 %agreeplot(data=table, var=Husband Wife, title=Husband and Wife Sexual Fun);

proc format;
     value rating 1="Certain" 2="Probable" 3="Possible" 4="Doubtful";
data msdiag;
     do Patients='Winnipeg   ', 'New_Orleans';
        do N_rating = 1 to 4;
           do W_rating = 1 to 4;
              input count @;
              output;
              end;
           end;
        end;
   format N_rating W_rating rating.;
   label N_rating = 'New Orleans neurologist'
         W_rating = 'Winnipeg neurologist';
datalines;
38  5  0  1
33 11  3  0
10 14  5  6
 3  7  3 10
 5  3  0  0
 3 11  4  0
 2 13  3  4
 1  2  4 14
;

%table(data=msdiag, var=Patients N_rating W_rating, weight=count, char=true, out=table);

%gdispla(OFF);
%agreeplot(data=table, var=Patients N_rating W_rating, by=patients, weights=FC2);
%gdispla(ON);

%panels(rows=1, cols=2, first=2);

See also