About Tom

Tom · ‎09-26-2011

Art - I think that the problem is that ViewTable does not distinguish between the formats that are defined in the dataset and those that it is using because of default behaviour. I would prefer that it just leave the format/informat attributes empty in that screen, as they are in the dataset. If they must put in a default value then it should be marked in someway to distinguish between that and the value of the attribute that is in the dataset's metadata.

Tom · ‎09-26-2011

proc freq ; tables response*category / norow nopercent ; run;

Tom · ‎09-25-2011

Looks like you want to loop over a set of SUFFIXES and use those suffixes for both specifing the name of the output dataset and the variable name. %macro stuff(suffixes); %local i j ; proc sql noprint; %do i=1 %to %sysfunc(countw(&suffixes)); %let j=%scan(&suffixes,&i); create table test&j as select role , cats(count(Q&j),'(',put(count(Q&j)/count(*),percent8.),')') as C&j from data group by role ; %end; quit; %mend stuff; %stuff(4 6 7 10 13);

Tom · ‎09-25-2011

Art - "Permanent formats" are just formats that are permanently attached because they are defined in the datasets. As opposed to formats that you temporarily attach during a procedure by using a format statement. - Tom

Tom · ‎09-25-2011

Are you sure you want a macro? If you are performing these operations in a datastep then just make an array and loop of the array. For example to recode 99 to missing for a select group of variables. array m99 q5-q12 q100-q112 ; do i=1 to dim(m99); if m99(i)=99 then varlist(i)=.; end; Now you cannot do the exact thing in macro language but you can get close by putting you list of variable names in one macro variable and assigning the variable name to another macro variable using a loop counter and the scan function. %let varlist = q5 q6 q7 q8 q9 q10 q11 q12 q100 q101 q102 q103 q104 q105 q106 q107 q108 q109 q110; %do i=1 %to %sysfunc(countw(&varlist)); %let var=%scan(&varlist,&i); . . . %end;

Tom · ‎09-25-2011

data want ; yeargap = 0; sizechange=0; do until (last.id) ; set have ; by id ; yeargap = max(yeargap , year ne lag(year) and not first.id); sizechange = max(sizechange, size ne lag(size) and not first.id); end; keep id yeargap sizechange; run;

Tom · ‎09-25-2011

This is one of my pet peeves with SAS's implementation of importing data from databases. They always attach both a format and an informat to character variables. As you have seen this can cause a number of headaches when you start combining that data with other datasets. In general permanently attaching the $ format to a character variable adds nothing of value. The only way I know to remove these is to use the FORMAT (or INFORMAT) statement with a variable list, but without a format. data sasfile ; set dbfile ; format _character_ ; run;

Tom · ‎09-25-2011

You seem to want to count the number of distinct values for a single variable (&COUNTVAR) when crossed with each of a number of other variables (&VARLIST). You can use the TYPES statement in PROC SUMMARY to generate a dataset that will make it easy to get those counts using PROC FREQ. The only problem I see is when your variables have missing values. %let varlist=level1 level2 ; %let countvar=featurenbr ; proc summary data=have nway ; class &varlist &countvar; types (&varlist)*&countvar ; output out=want1 ; run; proc freq data=want1 ; tables &varlist ; run;

Tom · ‎09-23-2011

What do you want as the output? You were concerned about getting only one number, but then you say that you only want one number for the variable FEATURE. If you look for distinct values of the variable FEATURE nested within the other two variables (see code above from FriedEgg) then only one combination has more than one distinct value of feature. If you want the number of distinct levels for each variable then try PROC FREQ with the NLEVELS option. That should tell you that the variable FEATURE has three distinct values.

Tom · ‎09-23-2011

On my Unix box I have to explicitely add the port number to the URL to read it from SAS code. Try adding :80 after the hostname. http://gdata.youtube.com:80/feeds/api/videos/pzI4D6dyp_o/comments?v=2&max-results=50&start-index=1

Tom · ‎09-23-2011

All of the sessions need use -RSASUSER. The first one has the PROFILE open so the second one cannot open it to write. With the -RSASUSER option SAS will not attempt to open it to write and it will work.

Tom · ‎09-23-2011

If the macro variable already exists the %GLOBAL does not change its value. In this piece of code the intent was to initialize the variable so that will have a value even if SQL did not find any observations. So the side effect would be that the pre-existing value persists even after this block of code runs. The %LET method does not care whether the variable is local (to the executing macro) or external (global or local to a surrounding macro scope) and so is more flexible. I do not like about overuse of %GLOBAL is that it clutters the global macro space. I write a lot of utility macros and I am careful to make sure to define the inputs and outputs to avoid unintended interactions. Within a macro I will frequently use %LOCAL to define (and initialize to empty) a variable.

Tom · ‎09-23-2011

You can prevent this by controlling your access to the SASUSER library. What I usually do is add the -RSASUSER option to my configuration file. This will make SAS open the SASUSER library in readonly mode and so it will not lock out the second (or third or fourth) session from accessing the PROFILE catalog. You can add this to either the configuration file or to the command line that you use to start SAS. The only trouble I have had is the occasional SAS tool that doesn't seem to understand that he RSASUSER option exists and is hard coded to write files to the SASUSER library. When you actually do want to modify something in your PROFILE then use the -NORSASUSER option. Another way is to create multiple directories with copies of your sasuser directory and use the -SASUSER option to point to different ones for different jobs. This has the disadvantage of requiring you to synchronize them.

Tom · ‎09-22-2011

What you experienced is most likely because the query did not match any observations. If you use SQL to create a macro variable and the query does not match any records then SQL does not create the macro variable. I usually just use a %LET before the query to set the value that I want if the query does not macth any observations. Using %global has other side effects, the worst being errors when the requested macro variable name already exists in a currently active local scope. proc sql noprint ; %let datebig=; select max(age) into : datebig from sashelp.class ; quit;

Tom · ‎09-20-2011

You should have also seen a message that the macro variable TABLE_LIST was not defined. This is because the query did not find any datasets that matched your query and so TABLE_LIST will never get defined. Most likely it is because of your use of lowercase strings. I know for sure that the LIBNAME value will always be uppercase. I think that the MEMNAME value will also always be in uppercase. Why not just start by listing out what tables SAS does show in the DICTIONARY.TABLES view? proc print data=dictionary.tables; var libname memname ; run; You also need to modify the catcatenation of libname and membername to eliminate trailing spaces on the LIBNAME variable. %macro merge_multiple; %let table_list=; proc sql noprint; select catx('.',libname,memname) into : table_list separated by " " from dictionary.tables where libname eq "WORK" and memname like 'DATA%'; quit; data work.want; %if &sqlobs %then %do; set &table_list; %end; run; %mend merge_multiple; data work.data1 ; ds=1; run; data work.data2; ds=2; run; %merge_multiple;

Online Status	Offline
Date Last Visited	Friday