I don't understand exactly what you are asking. What if any significance do the numbers following _B have?

Here is a bit of code that normalizies the data. Perhaps it will be helpful.

data work.qtr1; infile cards missover; input Uid:$1. x_b2001 y_b2001 z_b2001 x_B2051 y_b2051 z_b2051 x_b2301 y_b2301 z_b2301; cards; a 1 2 5 . . . 3 5 6 b 3 4 5 6 7 4 . . . c . . . 2 3 5 . . ;;;; run; data work.qtr2; infile cards missover; input Uid:$1. x_b2001 y_b2001 z_b2001 x_b2555 y_b2555 z_b2555 x_b2301 y_b2301 z_b2301; cards; a 1 2 5 . . . 3 5 6 b 3 4 5 . . . 4 7 9 c . . . 2 3 5 . . ;;;; run; data work.qtr / view=work.qtr; set work.qtr1(in=in1) work.qtr2(in=in2); quarter = in1*1 + in2*2; run;

proc transpose data = work.qtr out = work.Tqtr1(where=(not missing(col1))) ; by quarter uid; var x: y: z:; run; data work.Tqtr1; set work.Tqtr1; attrib xyz length=$1. label='Variable group (x,y,z)'; attrib num length=8 label='Number part of variable name'; xyz = _name_; num = input(substr(_name_,indexc(_name_,'bB')+1),f4.); run; proc print; run;

On 2/13/07, mesecca L katram <mesecca@yahoo.com> wrote: > Hi all > Here I have a large data set it has some 2000+ columns > the data is in the following format > Let us call this data set QTR_1 > > > Uid x_b2001 y_b2001 z_b2001 x_b2051 y_b2051 z_b2051 x_b2301 y_b2301 z_b2301 > a 1 2 5 . . . 3 5 6 > b 3 4 5 6 7 4 . . . > c . . . 2 3 5 . . . > > These is no pattern for the number for which x_b---- repeat but there 600 > sets of x_b---- y_b---- Z_b---- > > I have to creat a column count cnt_b2001 cnt_b2501 cnt_b2301 for each of > this and creat a data set > and then roll this up the resultant should look like this > > > x_b2001 y_b2001 z_b2001 cnt_2001 x_b2051 y_b2051 z_b2051 cnr_2051 x_b2301 > y_b2301 z_b2301 cnt_2301 > 4 6 10 2 8 10 9 2 3 5 6 1 > > This simply rolling up the data and counintg the non missing value > > > I have similar QTR_2, QTR_3 QTR_4 with almost 2000+ columns each > for which the variables are same but columns differ > > > QTR_2 > > Uid x_b2001 y_b2001 z_b2001 x_b2555 y_b2555 z_b2555 x_b2301 y_b2301 z_b2301 > a 1 2 5 . . . 3 5 6 > b 3 4 5 . . . 4 7 9 > c . . . 2 3 5 . . . > > > x_b2001 y_b2001 z_b2001 cnt_2001 x_b2555 y_b2555 z_b2555 cnt_2051 x_b2301 > y_b2301 z_b2301 cnt_2301 > 4 6 10 2 2 3 5 1 2 12 12 1 > > > > Finally I need to find the count > and display the report > Quater x y z count > QTR_1 > QTR_2 > QTR_3 > QTR_4 > > as X y Z patterns repaeat 600 times I cannot hardcode these > > What do you think is the most feasible way to do it > roll up by each qtr and then stack these the data might be in Giga bytes if > I stack before I roll upp > Please help > Thanks In advance >