Question

我有一个包含5000行和150个变量的sas数据集来自5000名受访者的调查，但是我需要删除整个行/响应者，其中列缺少对150个变量中的任何一个的观察。所以基本上，我只需要那些已完成150个变量答案的受访者。

我正在使用proc sql或base sas，但我无法想出更简单的方法来执行此操作。我使用了条件查询，但有些列是数字的，有些是字符类型而且我还需要对数字列进行分析，因此转置似乎不是替代.. 任何帮助将不胜感激？

由于

Answer 1

使用数据步骤只需：

data want;
  set have;
  if cmiss(of _all_) = 0;
run;

将处理字符和数字变量。

Answer 2

SAS procs倾向于通过从正在分析的数据中删除整行来忽略缺失值。所以，这可能不像你想象的那么严重。也就是说，如果您正在进行前向选择逻辑回归，请添加一组变量，然后只处理那些列没有缺失值的行。

如果要创建列没有缺失值的新数据集，可以执行以下操作：

proc sql;
    create table t_nomissing
        select t.* 
        from t
        where col1 is not null and col2 is not null and col3 is not null and . . .
              col150 is not null;

如果您有列名列表，我建议您在Excel等工具中创建where子句，您可以使用公式并将其复制下来。

Answer 3

让戈登·林诺夫（Gordon Linoff）关于Excel的想法更进一步只用SAS ......

ods output SQL_Results=appliance;
proc sql number;
select * from sashelp.applianc;
quit;


data appliance_2;
  set appliance;
  if cmiss(of _all_) = 0;
run;


proc sql; create table que as select * from dictionary.columns where libname = "WORK" and memname = "APPLIANCE"; quit;


proc sql ; 
select name, "IS NOT NULL AND"
from dictionary.columns where libname = "WORK" and memname = "APPLIANCE"; 
quit;

*copy / paste / clean-up ;
proc sql; 
create table appliance_3 as
select * from appliance 
where
Row IS NOT NULL AND 
units_1 IS NOT NULL AND 
units_2 IS NOT NULL AND 
units_3 IS NOT NULL AND 
units_4 IS NOT NULL AND 
units_5 IS NOT NULL AND 
units_6 IS NOT NULL AND 
units_7 IS NOT NULL AND 
units_8 IS NOT NULL AND 
units_9 IS NOT NULL AND 
units_10 IS NOT NULL AND 
units_11 IS NOT NULL AND 
units_12 IS NOT NULL AND 
units_13 IS NOT NULL AND 
units_14 IS NOT NULL AND 
units_15 IS NOT NULL AND 
units_16 IS NOT NULL AND 
units_17 IS NOT NULL AND 
units_18 IS NOT NULL AND 
units_19 IS NOT NULL AND 
units_20 IS NOT NULL AND 
units_21 IS NOT NULL AND 
units_22 IS NOT NULL AND 
units_23 IS NOT NULL AND 
units_24 IS NOT NULL AND 
cycle IS NOT NULL 
;quit;

PROC SQL / base SAS-从调查数据集中删除缺少的观察/没有响应

3 个答案: