That changes things then. A simple approach would be to use something like the scan function. newvar1=scan(orig_var, 1, '.'); newvar2=scan(orig_var, 2, '.'); etc ... View more

Specify the period as your delimiter. ... View more

Or if PROC SQL is your flavour (this is the 2nd post - accidentally submitted the first post and then deleted it): proc sql ; create table want as select cd.*   , coalesce ( q1.primarytable1, 0 ) as primarytable1   , coalesce ( q2.primarytable2, 0 ) as primarytable2   , coalesce ( q1.secondarytable1, 0 ) as secondarytable1   , coalesce ( q2.secondarytable2, 0 ) as secondarytable2 from /* create a set of unique codes from both datasets */ ( select distinct coalesce ( t1.code , t2.code) as code, coalesce ( t1.descrip, t2.descrip ) as descrip from table1 t1 full outer join ( select code, descrip from table2 ) t2 on t1.code = t2.code ) cd /* add up categories from table1 */ left join ( select code   , sum ( case when category='primary' then 1 else 0 end ) as primarytable1   , sum ( case when category='secondary' then 1 else 0 end ) as secondarytable1 from table1 group by code ) q1 on q1.code = cd.code /* add up categoris from table2 */ left join ( select code   , sum ( case when category='primary' then 1 else 0 end ) as primarytable2   , sum ( case when category='secondary' then 1 else 0 end ) as secondarytable2 from table2 group by code ) q2 on q2.code = cd.code order by code ; quit; ... View more

FYI - you can use a separate array for the regression formula. Alternatively, you can write the regression formula to a macro provided you use characters that can be evaluated by SYSEVALF. ... View more

So . . . you want to keep the original dataset (CLS), and you don't need the final dataset (CLS3)? ... View more

You could use an approach similiar to BallardW's suggestion but . . . Do you know if you will be "interpolating" over more than three data points (e.g. RateFA = (RateFA[i+1] - RateFA[i-1])/2 would apply if one rate was lower than the previous, but what if the second and third are both lower?). What if many subsequent data values for a given series are all lower than "expected"? What if the last value is lower? As an aside, it's unfortunate that your variables don't have a consistent naming pattern amenable to convenient array referencing (e.g. ARRAY RateFA {*} FA1-FA20). This forces you to explicity reference all of the variables in each array, as BallardW suggests. ... View more

Have you looked at the VNAME function? ... View more