基准数是基于主题访问（SAS）的SQL演示

Question

我正在尝试在特定的访问中计算基线平均值。例如，如果受试者没有进行指定的就诊，则将重新计算基线平均值，而忽略受试者。

这里是数据：

Subject            Visit              Value
001                Baseline           10
001                Visit 2            11
001                Visit 3            12
001                Visit 4            13
002                Baseline           11
002                Visit 2            12
002                Visit 4            13
002                Visit 5            14
003                Baseline           12
003                Visit 3            13
003                Visit 4            14
003                Visit 5            15

我想获得以下内容：

Visit             BaselineMean       VisitMean
Baseline          11                 11
Visit 2           10.5               11.5
Visit 3           11                 12.5
Visit 4           11                 13.3
Visit 5           11.5               14.5

这是我每次拜访时的平均值表：

proc sql;
create table want as
select 
    visit, 
    mean(value) as meanValue
from have
group by visit;

任何见识将不胜感激。

Answer 1

考虑两个聚合的联接，其中一个对自身使用自联接：

proc sql;
   CREATE TABLE want as 
   SELECT bagg.Visit, bagg.BaselineMean, vagg.VisitMean
   FROM 
     (SELECT t2.Visit, MEAN(t1.Value) AS BaselineMean
      FROM have t1
      INNER JOIN have t2
         ON t1.Subject = t2.Subject 
        AND t1.Visit = 'Baseline'
      GROUP BY t2.Visit) bagg

   INNER JOIN
     (SELECT Visit, MEAN(Value) AS VisitMean
      FROM have 
      GROUP BY Visit) vagg
   ON bagg.Visit = vagg.Visit;
quit;

SQL Demo

Answer 2

首先创建一个表，以具有每个主题的基线值，如下所示：

proc sql;
create table baseline as 
    select distinct subject, value
        from t1
    where visit = 'Baseline'
;

然后用基线值扩充主表。注意，如果没有主题的基线记录，则使用合并：

proc sql;
    create table inter as 
        select t1.*m, coalesce(b.value, 0) as b_val 
            from t1 left join baseline b  
                on t1.subject = b.subject
;
quit;

最终按以下方式计算基线和访问的平均值：

  proc sql;
        select visit, mean(b_val) as BaselineMean, mean(value) as visitMean
            from inter 
    group by visit
;
quit;