如何在SAS中将医疗CSV数据集的三分类Readmitted变量转为二分类？

No problem at all—here are two clean, reliable ways to recode your Readmitted categorical variable into a binary (0/1) format for logistic regression in SAS:

Method 1: Create a New Binary Variable (DATA Step) #

This approach makes a permanent new variable in your dataset, which is great if you'll use it across multiple analyses. Just adjust the dataset name (your_dataset) to match your actual data:

data cleaned_data;
    set your_dataset;
    /* Create binary variable: 0 = No readmission, 1 = any readmission */
    if Readmitted = "No" then Readmit_bin = 0;
    else if Readmitted in (">30", "<30") then Readmit_bin = 1;
    /* Optional: Handle missing values if needed */
    else Readmit_bin = .; /* Assign missing for any unexpected values */
run;

Method 2: Use a Custom Format (No New Variable) #

If you don't want to add a new variable, you can define a format that maps the original values to 0/1 on the fly. This works well for one-off analyses or preserving the original variable:

/* Define the custom format */
proc format;
    value readmit_bin_fmt
        "No" = 0
        ">30", "<30" = 1
        other = .; /* Map invalid values to missing */
run;

/* Apply the format in your logistic regression (or any procedure) */
proc logistic data=your_dataset;
    class Readmitted / param=ref; /* Optional, but useful for reference coding */
    model Readmitted = your_predictors / link=logit;
    format Readmitted readmit_bin_fmt.; /* Apply the binary format */
run;

Quick Notes: #

• Make sure the quotes around Readmitted values match the actual data type—if Readmitted is numeric instead of character, remove the quotes (e.g., if Readmitted = 0 then ... assuming "No" was coded as 0 originally).
• The other = . clause is optional but recommended to catch any unexpected values (like typos) and flag them as missing, which keeps your analysis clean.

内容的提问来源于stack exchange，提问作者Justin Arnold