Old macro ps from RAND

[pink_poodle]

Old macro ps from RAND corporation is trying to get SAS to play ping-pong with R. The macro is from a family called Twang, named after R package that gets propensity scores, hence the name ps. However, the variable names are illegal for SAS, for example, table.name is not a good variable name. I cannot even rename it, SAS complains. Workaround? I would like this code to work:

proc sql noprint;
select distinct table.name
into :elements separated by " "
from _baltab
order by table.name desc;
quit;

Many thanks!

[pink_poodle]

Must is come to this? I go into the preceding CSV file and manually alter table.name to table_name. Then I may run the remainder of the macro as a separate macro.

proc import datafile="&path/baltab.csv"
out=_baltab
dbms=csv
replace;
getnames=yes;
datarow=2;
guessingrows=MAX;

proc sql noprint;
select distinct table_name
into :elements separated by " "
from _baltab
order by table_name desc;
quit;

* %put &elements;
unw ks.max.ATT es.mean.ATT

[Tom]

Just set VALIDVARNAME option to V7 before running the PROC IMPORT and you will get a dataset with valid SAS variable names instead of names with periods and other invalid characters in them.

[pink_poodle]

Thank you, but this would also change the names of stat parameters in that
table. I just want to change table.name to table_name.

[Tom · Accepted solution]

So the other variables also have goofy names?

The RENAME statement or RENAME= dataset option can rename a variable name.

data want;
  set have;
  rename 'table.name'n=table_name;
run;

[pink_poodle]

I tried a literal. SAS would still complain. It does not like the dot. The funny thing is that csv with table.name variable is mostly for show. I can remove that part of the ps macro altogether.
It is pretty short, and most of the output is produced before.

[ballardw]

Log? Actual log contents? And since this apparently involves a macro then run with Options MPRINT.

You do not show exactly what SAS "complains" about. As in what the log says.

I don't think you are providing all the information.

Before you attempted a dataset name literal did you set Options Validvarname=any; ? That is required to reference data set variable name literals.

No complaints with this:

options validvarname=any;
data _baltab;
  'table.name'n='some garbage';
run;

proc sql noprint;
select distinct 'table.name'n
into :elements separated by " "
from _baltab
order by 'table.name'n desc;
quit;

%put elements are: &elements;

And the log:

249  options validvarname=any;
250  data _baltab;
251    'table.name'n='some garbage';
252  run;

NOTE: The data set WORK._BALTAB has 1 observations and 1 variables.
NOTE: DATA statement used (Total process time):
      real time           0.00 seconds
      cpu time            0.00 seconds


253
254  proc sql noprint;
255  select distinct 'table.name'n
256  into :elements separated by " "
257  from _baltab
258  order by 'table.name'n desc;
259  quit;
NOTE: PROCEDURE SQL used (Total process time):
      real time           0.03 seconds
      cpu time            0.00 seconds


260  %put elements are: &elements;
elements are: some garbage

So pretty much have to assume how ever you attempted to use a name literal was incorrect syntax.

[pink_poodle]

Woops, in the evening it works 🙂

proc import datafile="&path/baltab_dot.csv"
out=_baltab
dbms=csv
replace;
getnames=yes;
datarow=2;
guessingrows=MAX;

data test;
set _baltab;
rename 'table.name'n = table_name;
run;

[pink_poodle]

@ballardw you were right - incorrect syntax:

rename "table.name"/n = table_name;