Automating a process

[matt23]

Sorry for bad explanation but here's what I have:

I want a dataset with SLOPES for every hour. Now, I run this code:

proc reg data=have outest=outdata NOPRINT;
where Month in (3,4,5);
where also Weekday in (1,2,3,4,5);
model Load=Temperature DewPoint;
by Hour;
run;

proc reg data=have outest=outdata NOPRINT;
where Month in (3,4,5);
where also Weekday in (0,6);
model Load=Temperature DewPoint;
by Hour;
run;

and I get two tables that look similar to this:

Obs	Hour	_MODEL_	_TYPE_	_DEPVAR_	_RMSE_	Intercept	Temperature	DewPoint	Load
1	0	MODEL1	PARMS	Load	4121.22	72068.94	-440.384	328.518	-1
2	1	MODEL1	PARMS	Load	3840.58	70860.54	-491.293	341.535	-1
3	2	MODEL1	PARMS	Load	3619.64	70084.57	-509.326	327.479	-1
4	3	MODEL1	PARMS	Load	3455.33	70209.15	-550.968	338.977	-1
5	4	MODEL1	PARMS	Load	3388.95	70123.33	-545.392	315.663	-1
6	5	MODEL1	PARMS	Load	3348.44	71083.93	-566.492	316.459	-1
7	6	MODEL1	PARMS	Load	3326.66	73121.29	-581.957	307.265	-1
8	7	MODEL1	PARMS	Load	3399.56	76572.61	-670.97	371.085	-1
9	8	MODEL1	PARMS	Load	3491.99	76623.23	-544.325	300.572	-1

Now, I copy these two tables (weekend and weekdays) and play with them in EXCEL to make something like this:

Hour	WeekendDew	WeekdayDew
0	328.518	234
1	341.535	234
2	327.479	342
3	338.977	234

Is there a way to automate this somehow so I don't have to make my own table in EXCEL ?

[TomKari]

Easy peasy!

Create your first dataset as OutWeekDay, and your second as OutWeekEnd.

Then run

proc sql noprint;
	create table Combined as
		select E.Hour, E.Dewpoint as WeekendDew, D.Dewpoint as WeekdayDew
			from OutWeekEnd E join OutWeekDay D on
				E.Hour = D.Hour;
quit;

[PGStats]

Sure:

proc reg data=have outest=outdataDay NOPRINT;
where Month in (3,4,5);
where also Weekday in (1,2,3,4,5);
model Load=Temperature DewPoint;
by Hour;
run;

proc reg data=have outest=outdataEnd NOPRINT;
where Month in (3,4,5);
where also Weekday in (0,6);
model Load=Temperature DewPoint;
by Hour;
run;

data outDataDew;
merge
    outdataDay(rename DewPoint=weekDayDew) 
    outdataEnd(rename DewPoint=weekEndDew);
by hour;
keep hour weekDayDew weekEndDew;
run;

[ballardw · Accepted solution]

Send the output to different set and merge them while renaming the variable of interest:

proc reg data=have outest=outdata1 NOPRINT;
where Month in (3,4,5);
where also Weekday in (1,2,3,4,5);
model Load=Temperature DewPoint;
by Hour;
run;

proc reg data=have outest=outdata2 NOPRINT;
where Month in (3,4,5);
where also Weekday in (0,6);
model Load=Temperature DewPoint;
by Hour;
run;

data want;
   merge outdata1 (rename=(dewpoint=weekdaydew))
         outdata2 (rename=(dewpoint=weekenddew))
   ;
   by hour;
   keep hour weekenddew weekdaydew;
run;

I think should work. but without input data it is hard to verify.

[PGStats]

If you don't mind a different data structure:

data haveReg;
set have;
if month in (3,4,5);
if weekday in (0,6) 
    then dayEnd = "WeekEnd"; 
    else dayEnd = "WeekDay";
run;

proc sort data=haveReg; by dayEnd hour; run;

proc reg data=have outest=outdata NOPRINT;
by dayEnd hour;
model Load=Temperature DewPoint;
run;