About Tom

Tom · ‎09-26-2011

You just got some extra characters into your program. The error message is that it did not understand the PROC statement. So of course it does not understand the other statements that only apply when running PROC EXPORT.

Tom · ‎09-26-2011

Look at this thread. http://bbs.cdisc.org/bbs/forums/thread-view.asp?tid=3473

Tom · ‎09-26-2011

As Art says add PATIENT to your TABLES statement. If you want it as a list then use the LIST option. Otherwise you will get a grid with #rows equal to the number of patients and #cols equal to the number of categories. TABLES patient*diagnosis / list ;

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.

Online Status	Online
Date Last Visited	Monday