SAS中多变量组合打分的高效实现方案求助

Hey Rossco2210, great question—handling all those variable combinations manually is such a pain, especially with 10 variables where the number of possible combinations explodes exponentially. Let’s break down some cleaner, more efficient alternatives that avoid writing endless if/else blocks:

1. Group Similar Conditions with Logical Aggregation #

First, look for patterns in your scoring rules instead of listing every single combination. Your original code has multiple scenarios that result in a score of "4"—you can bundle those together using OR logic to cut down on repetition:

if (ai_hr_tat_flag = "Y" and ai_wtp_flag = "N" and ai_prof_flag = "N") /* Covers first two 4-score cases */
   or (ai_hr_flag = "Y" and ai_wtp_flag = "Y" and ai_prof_flag = "N") /* Third 4-score case */
then score = "4";
else if (ai_hr_tat_flag = "" and ai_hr_flag = "Y" and ai_wtp_flag = "N" and ai_prof_flag = "Y")
then score = "5";

This way, you’re grouping combinations that share core criteria instead of writing a separate else for each one. For 10 variables, this will drastically reduce the number of lines you need.

2. Use a Lookup Table for Exact Custom Combinations #

If your scoring rules are completely arbitrary (no clear patterns), a lookup table is your best bet. You can define all valid variable-score pairs in a separate dataset, then match your main data to it—no messy conditional logic required:

/* Step 1: Create a lookup table with all your required combinations */
data score_lookup;
   length ai_hr_tat_flag ai_hr_flag ai_wtp_flag ai_prof_flag $1 score $1;
   input ai_hr_tat_flag $ ai_hr_flag $ ai_wtp_flag $ ai_prof_flag $ score $;
   datalines;
Y N N N 4
Y Y N N 4
N Y Y N 4
  Y N Y 5  /* Empty string matches your "" case */
;
run;

/* Step 2: Merge your original data with the lookup table */
proc sql;
   create table scored_data as
   select a.*, b.score
   from your_original_dataset a
   left join score_lookup b
   on a.ai_hr_tat_flag = b.ai_hr_tat_flag
      and a.ai_hr_flag = b.ai_hr_flag
      and a.ai_wtp_flag = b.ai_wtp_flag
      and a.ai_prof_flag = b.ai_prof_flag;
quit;

For 10 variables, you just add rows to the lookup table for each combination you need to score—way easier to maintain and less error-prone than typing out every if/else.

3. Arrays Aren’t Scary—They’re Perfect for This! #

I get the hesitation, but arrays are made for handling batches of similar variables like your flags. They let you loop through variables without repeating code, which is a game-changer for 10 variables:

data scored_data;
   set your_original_dataset;
   /* Define an array to hold all your flag variables */
   array flags[4] $ ai_hr_tat_flag ai_hr_flag ai_wtp_flag ai_prof_flag;
   
   /* Example: Count how many flags are set to "Y" */
   count_y = 0;
   do i = 1 to dim(flags);
      if flags[i] = "Y" then count_y + 1;
   end;
   
   /* Use the count (or other checks) to assign scores */
   if (count_y = 1 and ai_hr_tat_flag = "Y") or (count_y = 2 and ai_wtp_flag = "Y")
   then score = "4";
   else if count_y = 2 and ai_hr_tat_flag = "" and ai_prof_flag = "Y"
   then score = "5";
run;

With 10 variables, you just update the array definition to include all 10 flags—no need to rewrite variable names over and over. It’s way cleaner than endless conditionals.

Pick the approach that fits your scoring rules best: logical grouping for pattern-based rules, lookup tables for arbitrary combinations, and arrays for batch variable operations.

内容的提问来源于stack exchange，提问作者Rossco2210