About ArtC

ArtC · ‎08-13-2012

[SAS 9.3] Here is a format which uses an FCMP function to split the incoming string and calculate the fraction. It also uses the FRACT. format to write the numeric result. Notice one of the odd results of the FRACT format - something for me to look into. options cmplib=work.functions; proc fcmp outlib=work.functions.myfunc; function nfract (f$) ; return(input(scan(f,1,'/'),best.)/input(scan(f,2,'/'),best.)); endsub; run; proc format; invalue $readfrac other = [nfract()]; run; * Show two ways to convert the value; data have; input cfract $8. @1 xfract $readfrac8.; put cfract= ; put xfract; nfract= nfract(cfract); put nfract= nfract= fract10.; datalines; 1/4 234/67 3/4 run;

ArtC · ‎08-13-2012

It is not a matter of importing the value, but rather one of storage. SAS uses a length of 8 to store numerics, this allows 16 digits to be stored accurately (depending on your OS). See http://support.sas.com/techsup/technote/ts654.pdf for more information. So what to do? It depends. What do you want to do with the data value? Arithmatic? Is it a code that can be converted to character as suggested above?

ArtC · ‎08-11-2012

What I see is a double conversion using three formats. Since we don't see your data I will guess that it looks something like: proc format; value $vxxaa 'a' = '1' 'b' = '2' 'c' = '3'; value vxx 1 = 10 2 = 20 3 = 30; invalue vxxbb 'a' = 1 'b' = 2 'c' = 3; run; data have; input vxx $; newvar = input(put(vxx,$vxxaa.),best.); altvar = input(vxx,vxxbb.); put newvar= vxx.; put altvar= altvar vxx.; datalines; a b c run; Notice here that the informat VXXBB can be used to simplify the INPUT/PUT combination. We could simplify further if the variable NEWVAR could contain its formatted value. As to your errors it may be because VXX is not initially character.

ArtC · ‎07-31-2012

Good catch Haikuo, This one has a couple fixes. data _null_; if 0 then set have; dcl hash h(dataset:'have', ordered: 'a', multidata:'Yes'); h.definekey('entity'); h.definedata ('entity','info1','info2'); h.definedone(); dcl hiter hh('h'); length value $8; value=' '; do while(hh.next()=0); if value ne entity then rc=h.output (dataset:compress(entity,,'kda')||'(where=(entity="'||entity||'"))'); value=entity; end; stop; run;

ArtC · ‎07-31-2012

haikuo, Good idea to use the hash object. The following alternate approach utilizes the HITER object and IMO simplifies the DATA step a bit. data _null_; if 0 then set have; dcl hash h(dataset:'have', ordered: 'a'); h.definekey('entity'); h.definedata ('entity','info1','info2'); h.definedone(); dcl hiter hh('h'); do while(hh.next()=0); rc=h.output (dataset:compress(entity,,'kda')); end; stop; run;

ArtC · ‎07-31-2012

For some reason we often make SAS dates more complicated than necessary when working with them in the macro language. Remember that a date value is just a number. The date constant form such as '31Jul2012'd is actually a function call that is translated to a number, 19205. Therefore there is no reason to convert a SAS date to a text string and then convert it back to a number. For instance: %let test_end_date = %sysfunc(intnx(month, "&sysdate"d, &shift, e), date9.); For today's date the value of &test_end_date is 31jul2012. and when used in the constant "&test_end_date"d resolves to 19205. We can simplify by going directly to the actual date. The %LET becomes: %let test_end_date = %sysfunc(intnx(month, "&sysdate"d, &shift, e)); &test_end_date now contains 19205 and it can be used directly in a comparison: %let date_string = &test_start_date <= date <= &test_end_date;

ArtC · ‎07-31-2012

First the easy part: Let's simplify the creation of the macro variables that hold the dates. Call symputx("Test_Start_Date", intnx("Month", Today, -&i); Notice that I used the SYMPUTX routine and eliminated the 'B' (the default). Second The WHERE clause in your SQL step depends on the macro loop so why would you want it to only execute once? I think that I do not understand. Perhaps you want to write the code once and execute it twice - which is what you have done here.

ArtC · ‎07-26-2012

Let us know how it works for you.

ArtC · ‎07-24-2012

In your revised problem I will assume that the column NI is the mean of the NI values. This problem is a form of a running average problem. I will tackle it here using a multlabel format - there are other approaches. In the following code substitute SEX AGE and WEIGHT for your variables the approach will be the same. proc sort data=sashelp.class out=have; by sex age; run; proc print data=have; run; proc format ; value yrgrp (multilabel) 10 = '10' 10-11 = '10-11' 10-12 = '10-12' 10-13 = '10-13' 10-14 = '10-14' 10-15 = '10-15' 10-16 = '10-16' 10-17 = '10-17' 10-18 = '10-18'; run; proc summary data=have; class sex; class age/mlf; var weight; ways 2; output out=groups n= mean= std=/autoname; format age yrgrp.; run; proc print data=groups; run;

ArtC · ‎07-24-2012

Are you still getting missing values for the STD? My test code gives the expected results.

ArtC · ‎07-23-2012

If we can assume that the SQL step was actually stated something like: proc sql; _______________ from list.size,list.price where size.address =price.address; I would have also selected D.

ArtC · ‎07-23-2012

Are you sure about the SQL step code? The semi-colon after the WHERE clause will cause the FROM to be separated from the WHERE.

ArtC · ‎07-23-2012

Silly question but can you verify that there are at least two values of NI for each combination of your GROUPBY variables, and that NI is not missing for at least two obs within the BY group?

ArtC · ‎07-23-2012

Thanks Ksharp - good catch. Or the WAYS statement which offers more flexibility.

ArtC · ‎07-23-2012

As an aside notice in Cynthia's code that the compute block is for MAXWEIGHT and not WEIGHT. In the original posted code you are attempting to use the variable MAXWEIGHT in the WEIGHT compute block. Since MAXWEIGHT is to the right of WEIGHT in the COLUMN statement, MAXWEIGHT cannot yet be addressed. Remember that for a compute block associated with a given report item you may only address report items to the left of it in the COLUMN statement.

Online Status	Offline
Date Last Visited	‎06-20-2021 07:11 PM