使用sas识别数据集中的相关值对

Question

我有一个包含单词同义词信息(多行)的数据集，下面是数据集的简要示例。给出了每个单词的同义词信息。

Word Synonym
C01  C02
C01  C05 
C02  C02
C02  C05
C03  C04
C05  C06
C11  C12
..   ..

从上面的数据集中，单词-同义词的关系可以识别如下。

C01-C02-C05-C06

C03-C04

C11-C12

执行完sas代码后，我想要一个如下所示形式的数据集。

Word Synonym1 Synonym2 Synonym3
C01  C02      C05      C06
C03  C04
C11  C12

我尝试了内部连接的冗余步骤，但似乎有很多不必要的过程。

Answer 1

我很难在SAS中找到一个好的解决方案(在其他语言中，这更容易解决)。下面的方法不是很好，因为它试图将所有组写入单个变量中，如果您有很多记录，这个变量很快就会用完。另外，它依赖于'#‘作为分隔符。如果您的单词可以包含此字符，您可能希望将其更改为不同的字符。

data groups;
    set testData nObs=numObs;

    array groups [*] $32767 group1-group100;
    retain groupN 0 group1-group100;

    categorized = 0;

    * Search for the word or synonym in the existing groups;
    if (groupN >= 1) then do;
        do currentGroup = 1 to groupN;
            if (index(groups[currentGroup], "#"||strip(word)||"#") and index(groups[currentGroup], "#"||strip(synonym)||"#") = 0) then do;
                    groups[currentGroup] = strip(groups[currentGroup])||strip(synonym)||"#";
                    categorized = 1;
            end;
            if (index(groups[currentGroup], "#"||strip(word)||"#") = 0 and index(groups[currentGroup], "#"||strip(synonym)||"#")) then do;
                    groups[currentGroup] = strip(groups[currentGroup])||strip(word)||"#";
                    categorized = 1;
            end;
            if (index(groups[currentGroup], "#"||strip(word)||"#") and index(groups[currentGroup], "#"||strip(synonym)||"#")) then do;
                    categorized = 1;
            end;

        end;
    end;

    * If the word and synonym were not found in the existing groups, create a new one;
    if (categorized = 0) then do;
        groups[groupN + 1]  = "#"||strip(word)||"#"||strip(synonym)||"#";
        groupN = groupN + 1;
    end;

    * Split the groups into unique key/value pairs;
    if (_n_ = numObs) then do;
        length key value $200;
        keep key value;
       do currentGroup = 1 to groupN;
            if (not missing(groups[currentGroup])) then do;
                key = scan(groups[currentGroup], 1, '#');
                do j = 2 to countC(groups[currentGroup],'#');
                    value = scan(groups[currentGroup], j, '#');
                    if (not missing(value)) then do;
                        output;
                    end;
                end;
            end;
       end;
    end;
run;

proc sort data = groups;
    by key;
run;

proc transpose data = groups out=result(drop = _:) prefix=synonym;
    by key;
    var value;
run;