SAS MACRO SUMMARIZING VARIABLE INFORMATION

shiyiming · 发表于 2010-10-22 13:24:08

From Wensui Liu's blog

%macro summ(data = ); ****************************************************; * THIS MACRO IS TO SUMMARIZE VARIABLES INFORMATION *; * IN A SAS DATA SET *; ****************************************************; options mprint mlogic orientation = landscape nodate nocenter nonumber; proc contents data = &data varnum out = _meta noprint; run; data _nmeta (rename = (index1 = i)) _cmeta (rename = (index2 = i)); set _last_ end = eof; if type = 1 then do; index1 + 1; output _nmeta; end; else if type = 2 then do; index2 + 1; output _cmeta; end; run; proc sql noprint; select max(i) into :num_cnt from _nmeta; select max(i) into :cha_cnt from _cmeta; select count(*) into :cnt from &data; quit; %local num_cnt cha_cnt cnt; %do num_i = 1 %to &num_cnt; proc sql noprint; select strip(upcase(name)) into :nx from _nmeta where i = &num_i; select n(distinct &nx) into :dist from &data; quit; %local nx dist; proc summary data = &data nway; output out = _data_ (drop = _freq_ _type_) n(&nx) = cnt nmiss(&nx) = mis min(&nx) = xmin p25(&nx) = xp25 median(&nx) = xmdn p75(&nx) = xp75 max(&nx) = xmax mean(&nx) = xavg; run; %if &num_i = 1 %then %do; data _nsumm; length var $32; set _last_; var = "&nx"; val_pct = cnt / &cnt; mis_pct = mis / &cnt; dis = &dist; run; %end; %else %do; data _nsumm; set _nsumm _last_ (in = new); if new then do; var = "&nx"; val_pct = cnt / &cnt; mis_pct = mis / &cnt; dis = &dist; end; run; %end; %end; %do cha_i = 1 %to &cha_cnt; proc sql noprint; select strip(upcase(name)) into :cx from _cmeta where i = &cha_i; create table _level1 as select &cx as level, count(*) / &cnt. as pct from &data group by &cx order by pct descending; select count(*) into :level from _level1; select case when compress(level, ' ') = '' then '_' else strip(level) end ||"("||strip(put(pct, percent5.0))||")" into :tops separated by ', ' from _level1(obs = 5); quit; %local cx tops; %if &cha_i = 1 %then %do; data _csumm; length var $32 tops $100; var = "&cx"; level = &level; tops = "&tops"; run; %end; %else %do; data _csumm; set _csumm end = eof; output; if eof then do; var = "&cx"; level = &level; tops = "&tops"; output; end; run; %end; %end; title; %if &num_cnt > 0 %then %do; proc report data = _nsumm box spacing = 1 split = "*"; column("SUMMARY OF NUMERIC VARIABLES (obs = %trim(&cnt))" var val_pct mis_pct dis xmin xp25 xmdn xp75 xmax xavg); define var / "VARIABLE" width = 20 order order = data center; define val_pct / "VALID*PERCENT" width = 10 format = percent9.2; define mis_pct / "MISSING*PERCENT" width = 10 format = percent9.2; define dis / "DISTINCT*VALUE" width = 10 format = 9.; define xmin / "MININUM" width = 10; define xp25 / "1ST*QUANTILE" width = 10; define xmdn / "MEDIAN" width = 10; define xp75 / "3RD*QUANTILE" width = 10; define xmax / "MAXIMUM" width = 10; define xavg / "MEAN" width = 10; run; %end; %if &cha_cnt > 0 %then %do; proc report data = _csumm box spacing = 1 split = "*"; column("SUMMARY OF CHARACTER VARIABLES (obs = %trim(&cnt))" var level tops); define var / "VARIABLE" width = 20 order order = data center; define level / "# of*LEVELS" width = 10; define tops / "TOP 5*CATEGORIES" width = 100 center; run; %end; %mend summ;

		自动登录	找回密码
密码			立即注册