The +(-1) is tell it to back over the extra space that it normally puts between variables when using this list mode style of put statement.  The +(-1) is not needed when using DSD option.  You do not want the cursor movement command in your delimiter, so you would want to set DLM='09'x in the FILE statement.  (or you could use dlm=&tabf as it would be the same thing). The use of &TABF is to have it put out two tabs with no space between them.  To mimic that with DSD option you will need to make an empty char variable. blank=' '; PUT RECORDNO FIPS STATEAGN DISTRICT blank ImpDS EXPLAIN; ... View more

Actually looking at what your code is doing you should probably not be including the denominator variable in the array.  Then it would not matter if it was before, after or in the middle of the list of numerator variables in the data vector. data share_three;                                                                                                     set &cleaned_data(keep=store_id year quarter vol_g: vol_i: );     array share vol_g: ;                                                                     do over share;                                                                                            if vol_i_0000 = 0 then share=0;     else share = share / vol_i_0000 ;   end;   * Un comment next line if you want to force VOL_I_0000 to 1 as your previous code did ;   * if vol_i_0000 ne 0 then vol_i_0000=1; run; ... View more

Probably this code is just a simplified version of what you are really doing, but it does not look like an application for a macro arithmetic.  Seems it would be more appropriate to just write it in SAS rather then using the macro language to generate SAS code. data doudou;   i=1;   j=2;   set mydata;   array a a:;   a((i-1)*18+(j-1)*6)=1; run; Or perhaps you need a two dimension array? data doudou;   i=1;   j=2;   set mydata;   array a(18,6) a:;   a(i,j)=1; run; ... View more

I assume from the macro variable names the intent is a TAB delimited file. Note that with the addition of the DSD and DLM options to the FILE statement you could probably get rid of the &TAB macro variable references.  But what the purpose of &TABF macro variable is I can only guess perhaps to have any empty column?  filename tempmk dsd dlm="&tab"; ... PUT FILETYPE RECORDS FILENAME FILEID FILEYEAR ' '; .... PUT RECORDNO FIPS STATEAGN DISTRICT &TABF ImpDS EXPLAIN; .... ... View more

You need to specify lengths for your data file records and your variables. Add LRECL option to your INFILE statement. Increase the length of the TEMP _TEMP variables. If you have 21 variables @ 4000 characters per variable then you might need set the length to 85000.  That would work for LRECL option, but is too long for a data step variable.  Most likely any given record will not have maximum length for each variable so you could try setting length of the temp strings to 32767. Let me emphasize again that the real issue here is the process that is generating the file.  If you have any control over that then you can eliminate this problem by fixing the way that the data file is written. ... View more

If you are using PROC SQL you can use the CREATE TABLE clause to create a table (dataset). proc sql ;    create table WANT as      select x,y,y*max(x) as z      from HAVE   ; quit; You should also learn how to use SAS Procs to product output tables. proc summary data=HAVE ;    var x;    output out=xmean mean=xmean ; run; And how to manipulate data using DATA steps.  Here is a trick to copy values from a dataset with only one observation onto all observations of another dataset.  _N_ is the automatic loop counter for the data step.  So the record from the dataset XMEAN will only be read the first time SAS runs through the data step code.  Since XMEAN variable is never overwritten again it will stay the same for all iterations of the data step. data want;   set have ;   if _n_=1 then set xmean;   z=y*xmean; run; ... View more

Most likely this is being caused by PROC IMPORT not being able to detect the type and length of the fields. Look up the quessing rows option. I could not even get PROC IMPORT to read your Assassins_Creed.csv file.  Most likely because of the UTF-8 warnings that I was getting before.  The UTF-8 issues might also be causing the trouble with the negative sign.  Perhaps the character that is actually in the file is something other than a hyphen and it just looks to your eye like a negative sign?  If so then SAS will not know how to interpret that character. 2839  filename xx "&fname" encoding='utf-8'; 2840  PROC IMPORT OUT= WORK.test1 2841              DATAFILE= xx 2842              DBMS=CSV REPLACE; 2843       GETNAMES=NO; 2844       DATAROW=3; 2845  RUN; ERROR: Invalid string. ERROR: Invalid string. ERROR: Import unsuccessful.  See SAS Log for details. NOTE: The SAS System stopped processing this step because of errors. That is why it is easier for you to just write a DATA step to read these files.  You can use the names pulled from the first line to generate the code yourself. proc sql noprint ;   select name into :keys separated by ' ' from names where col <=12 ;   select name into :questions separated by ' ' from names where col > 12; quit; data results ;   infile "&fname" dsd dlm=',' truncover firstobs=3 lrecl=500000;   length &keys $100 &questions $4 ;   input &keys &questions ; run; ... View more

I looked at the file "Assassins_Creed.csv" . You have 732 (!!!!) columns. The second line with the question values is over 100,000 characters long. The file is using UTF-8, but there is some strange character in there that SAS cannot transcode into English.  Perhaps it is the strange symbol that I see at the end of the second line? Not sure that there is going to be any easy way to deal with this format.  There appear to be three types of rows in that file.  The first has (almost) variable names.  The second is a mix of variable names (or perhaps labels) for the first 12 columns and then information about the question asked for the rest of the columns. The string here looks a little like it also might have other information about the question as a number of the values appear to be multiple values separated by hyphens. Then the remaining rows appears to have the actual data.  The first 12 columns could perhaps be read using defined variable types. The rest seem to have coded values (even though names end in 'TEXT').  The FREQ Procedure                                      Cumulative    Cumulative response    Frequency     Percent     Frequency      Percent ------------------------------------------------------------- -1                43       20.28            43        20.28 -2                10        4.72            53        25.00 0                 22       10.38            75        35.38 1                 39       18.40           114        53.77 2                 16        7.55           130        61.32 3                  1        0.47           131        61.79 N/A               81       38.21           212       100.00 Here is an attempt to read this file into three datasets in a vertical format. *-----------------------------------------------------------------------------; * Read in response file ; *-----------------------------------------------------------------------------; %let fname=c:\downloads\transpose\Assassins_Creed.csv; options compress=yes; *-----------------------------------------------------------------------------; * Get the variable names from first row. Clean up to valid SAS variable names ; *-----------------------------------------------------------------------------; data names ;   infile "&fname" dsd dlm=',' truncover lrecl=500000 obs=1;   length col 8 name $32 label $256;   do col=1 by 1 until(name=' ');      input label @;      name = translate(label,'_','#');      if name ne ' ' then output;   end;   call symputx('NCOL',col-1); run; %put Found &ncol column names; *-----------------------------------------------------------------------------; * Get the question names from second row. ; *-----------------------------------------------------------------------------; data questions ;   infile "&fname" dsd dlm=',' truncover lrecl=500000 obs=2 firstobs=2;   length col 8 question $1000 ;   do col=1 to &ncol;      input question @;      if index(question,'Click to write the question text') then        question=left(substr(question,index(question,':')+1))      ;      if question ne ' ' then output;   end; run; /* proc report data=questions nofs split='FF'x nocenter headline missing spacing=1; column col question; define col / order ; define question / display flow width=75 ; run; */ *-----------------------------------------------------------------------------; * Get the response from the third row and forward ; *-----------------------------------------------------------------------------; data responses ;   infile "&fname" dsd dlm=',' truncover lrecl=500000 firstobs=3;   length row col 8 response $1000 ;   row+1;   do col=1 to &ncol;      input response @;      if response ne ' ' then output;   end; run; ... View more