About supersonic

sandyming · ‎08-23-2013

proc stdize data=old out=new missing=mean reponly; var _numeric_; run; You can also calculate the median first, then read values from a second data set

supersonic · ‎08-21-2013

Thank you Astounding and Murray_Court

AncaTilea · ‎08-05-2013

ok, it's a bit hard for me to understand the meaning of "mode" for categorical variables. So, what about this: DATA TEMP; SET SASHELP.CLASS; IF NAME = "Janet" THEN NAME = "Jane"; IF NAME = "James" THEN NAME = "Jane"; IF . <AGE < 12 THEN AGE_GRP = "< 12 yrs "; ELSE IF 12 <= AGE <= 13 THEN AGE_GRP = "12-13 yrs"; ELSE AGE_GRP = "> 13 yrs"; RUN; PROC FREQ DATA = TEMP; TABLES SEX AGE_GRP; ODS OUTPUT ONEWAYFREQS = MY_F(KEEP = F_: FREQUENCY); RUN; PROC UNIVARIATE DATA = MY_F; VAR FREQUENCY; WHERE F_AGE_GRP NE " "; RUN; PROC UNIVARIATE DATA = MY_F; VAR FREQUENCY; WHERE F_SEX NE " "; RUN; In the first step, I modify the data SASHELP.CLASS to create another categorical variable, age_grp. Then you run a frequency for the categorical variables...output a data set that will give you no:of distinct categories. Then run a univariate on it...and you get the mode (if it exists)...

ballardw · ‎08-05-2013

You mention that your data only has two variables one binary and the other the predicted probability. Since one of your requirements is for the average predicted probability then what do want a decile of? The binary only has two levels and making a decile from that is going to be a tad rough. Also, what would be the differece be number of observations and events. I think you need to describe your starting data a bit more fully.

Tom · ‎06-26-2013

How to you want the information formatted? You could put them into those same three macro variables as delimited values. select memname , nvar , nobs into :m1 separated by ' ' , :v1 separated by ' ' , :o1 separated by ' ' Or you could generated multiple variables. select memname , nvar , nobs into :m1 - :m999999 , :v1 - :v999999 , :o1 - : o999999 ... %let n=&sqlobs; There is one difficulty with doing this as a macro since by default to new macro variables will be local to the macro and hence not available when the macro ends. You could force them into the GLOBAL symbol table. In which case it might be easier to do using CALL SYMPUTX. data _null_; if eof then call symputx('n',_N_-1,'G'); set sashelp.vtable end=eof; where libname = %upcase("&lib") ; call symputx(cats('m',_n_),memname ,'G') ; call symputx(cats('v',_n_),nvar ,'G') ; call symputx(cats('o',_n_),nobs ,'G') ; run;

supersonic · ‎06-24-2013

thank you Ksharp and SAS333 %macro lib(lib=); proc sql noprint; select memname,nvar,nobs into : m1, : v1, : o1 from dictionary.tables where libname="%upcase(&lib)" and memtype="DATA"; quit; %mend lib; %lib(lib=)

SteveDenham · ‎06-21-2013

What about the standard BASE procs: MEANS and FREQ? Steve Denham

Astounding · ‎06-20-2013

You can use %BY, but you can't use a decimal fraction like 0.1. The last construction where DO can be used but %DO cannot is a combination of a range with WHILE or UNTIL. Legitimate: do i=1 to 10 until (age > 21); No good: %do i=1 %to 10 %until (&age > 21); You can use %DO %WHILE or %DO %UNTIL, you just can't combine that with a range of values.

Linlin · ‎06-13-2013

%macro reg(dsname, DV, PV, selecttype ); proc logistic data=&dsname; model &DV=&PV /selection = &selecttype; output out=out_&dsname ; run; title from dataset out_&dsname; print data=out_&dsname; run; title; %mend reg;

Online Status	Offline
Date Last Visited	‎09-01-2015 07:12 AM

Re: replace missing value with mean

replace missing value with mean

Re: macro help

Re: macro help

Re: macro help

Re: macro help

macro help

Re: univariate

univariate

Re: sas macro

Re: Do but not %do

Re: replace missing value with mean

Re: macro help

Re: univariate

Re: sas macro

Re: no:of obs & vars

Re: SAS MACRO

Re: analysis macro >> imp

Re: Do but not %do

Re: SAS MACRO