topic From Stata to SAS Code in Statistical Procedures

From Stata to SAS Code

Artp — Fri, 24 Nov 2017 15:18:24 GMT

Hello Friends,

I am not STATA user. I tried to understand this language so I could translate its code into SAS. I need your input in certifying that my SAS code is equivalent to the Stata code.

/*Example STATA code:*/

gen strataid=.
replace strataid=(GESTFIPS*10000)+ HG_MSAC if HG_MSAC > 0
replace strataid =(GESTFIPS*1000) + GECO if GECO > 0 & HG_MSAC==0
replace strataid = GESTFIPS if GECO == 0 & HG_MSAC == 0

This is how I converted the above code into SAS:
................................;
length STRATAID 4;
set stratumxplr;

if missing(GECO) then call missing (STRATAID);
if missing (HG_MSAC) then call missing (STRATAID);
if missing (GESTFIPS) then call missing (STRATAID);
if GTCBSA > 0 then STRATAID = (GESTFIPS*10000) + GTCBSA;
else if GTCO > 0 and GTCBSA = 0 then STRATAID = (GESTFIPS*1000) + GTCO;
else if GTCO = 0 and GTCBSA = 0 then STRATAID = GESTFIPS;
label STRATAID = 'Stratum Identification';
run;

Are the Stata double == equal signs equivalent to SAS one equal sign?

Stata code is from Michael Davern et al. p. 14 (http://journals.sagepub.com/doi/pdf/10.5034/inquiryjrnl_43.3.283).

Thanks for your help,

Artp

====

Re: From Stata to SAS Code

Kurt_Bremser — Fri, 24 Nov 2017 20:48:11 GMT

If stata follows C rules (likely), == is the comparison operator, while = assigns values. In SAS, both are done with a single =.

Re: From Stata to SAS Code

Kurt_Bremser — Fri, 24 Nov 2017 20:54:37 GMT

PS for final verification, practice beats theory: run both codes in their respective environments and compare the results.

Re: From Stata to SAS Code

Tom — Sat, 25 Nov 2017 04:27:42 GMT

It looks like the STATA code is unconditionally initiating the new variable to missing.

It does not look like the STATA code is using anything that is the equivalent of the ELSE statements in your code. But since the conditions look to be mutually exclusive it shouldn't matter.

Why did use set the LENGTH of your new variable to 4? SAS uses 64bit floating point numbers for all numeric variables. Storing only 4 of the 8 bytes will potentially remove some of the precision of the values.

In SAS equality testing is done with a single equal sign. Or the EQ keyword.