SAS宏与宏变量报错解决：%eval函数中字符操作数问题及变量移位需求

Let's break down what's going wrong here and fix it properly.

Why the Error Happens #

The key issue is a fundamental mix-up between macro variables/processing and data step variables:

• %eval() is a macro function that runs during the compilation phase, before your data step executes. It only works with macro variables or literal values, not data step variables like diff.
• When you write %eval(&i. - diff), the macro processor sees &i as a numeric macro variable (1-5), but diff is just plain text to it—it has no idea this is a data step variable with numeric values. So it tries to subtract the string "diff" from a number, which triggers that "character operand in %eval" error.

Better Solution: Use SAS Arrays (No Macro Needed!) #

For this kind of repetitive variable manipulation (shifting values across d1-d5 to n1-n5), SAS arrays are far simpler, more efficient, and avoid macro/data step confusion entirely. Here's how to rewrite your code:

/* Create the example data (fixed initial data step syntax) */
Data example_table;
 input Name $ d1 d2 d3 d4 d5 diff;
 datalines;
A 0.2 0.2 0.1 0.2 0.3 1
B 0.3 0.1 0.4 0.3 0 2
C 0.1 0.2 0 0.4 0.3 2
;
run;

/* Generate the new shifted variables using arrays */
Data new_table;
 Set example_table;
 /* Define arrays for original d variables and new n variables */
 array d[5] d1-d5;
 array n[5] n1-n5;
 
 do i = 1 to 5;
   if i <= diff then do;
     n[i] = 0;
   end;
   else do;
     /* Shift by diff: take the d value at position (i - diff) */
     n[i] = d[i - diff];
   end;
 end;
 drop i; /* Clean up the loop variable */
run;

What This Does: #

1. Arrays: array d[5] d1-d5 creates a reference group for your 5 distribution variables, so you can access them by index (d[1] = d1, d[2] = d2, etc.). Same for n[5] for the new variables.
2. Loop: The do i = 1 to 5 loop runs once per variable index. For each index:
   • If the index is ≤ diff, set the corresponding n variable to 0.
   • Otherwise, pull the value from the d array at position i - diff (exactly the shift logic you need).
3. This code will produce exactly the new_table you showed in your example.

If You Insist on Using a Macro (Not Recommended Here) #

If you want to stick with a macro approach, you need to avoid using %eval() with data step variables. Instead, build the variable references dynamically in the data step using character functions:

%macro distr ();
 %local i;
 %do i = 1 %to 5;
   if &i. <= diff then n&i. = 0;
   else do;
     /* Use data step functions to build the d variable name */
     length d_var $3;
     d_var = cats('d', put(&i - diff, 1.));
     n&i. = input(vvaluex(d_var), best12.);
   end;
 %end;
%mend;

Data new_table;
 Set example_table;
 %distr();
 drop d_var;
 run;

But again, arrays are the standard SAS way to handle this scenario—they're cleaner, faster, and less prone to errors like the one you hit.

内容的提问来源于stack exchange，提问作者vicky