By default, PROC SUMMARY produces summaries for all combinations of the class variables, and also a grand total. If you specify NWAY, you will get only the lowest-level of combinations (highest level of detail). Since you have only one class variable, all NWAY does in this case is eliminate the total row.

MISSING says "keep observations with a missing value for one of the class variables". The default (in my opinion a bad choice) is to throw away those observations without warning. If you don't having any missing values, this has no effect, but I think it's good practice to always include it.

SUM= says that the kind of summariztion that you want is addition.

The _TYPE_ variable is used to distinguish the different levels of summarization. If NWAY is specified, _TYPE_ carries no information. If you had not specified NWAY, there would been two values of _TYPE_ in the output data set: 0 in the one row containing the grand total, and 1 in the detail rows.

-- JackHamilton@FirstHealth.com Manager, Technical Development Metrics Department, First Health West Sacramento, California USA

>>> "Michelle Jellinghaus" <michelle@emode.com> 06/19/2003 10:31 AM >>> This worked beautifully. Thank you very much for taking the time to answer my post.

I am still trying to figure out *why* it worked, however ;) I hope you won't mind a few questions.

This is what I have so far; will you let me know if I'm right? nway: this option removes an observation that contains a column sum for all rows missing: I'm not sure what this does here, but it seems to work :) sum= in the output statement: this seems to give a sum of all values in a column by class instead of returning M, STD, etc.

I see the _type_ column, but I don't understand what it is.

Thanks again for your help.

M

-----Original Message----- From: Jack Hamilton [mailto:JackHamilton@firsthealth.com] Sent: Wednesday, June 18, 2003 6:11 PM To: Michelle Jellinghaus; SAS-L@LISTSERV.UGA.EDU Subject: Re: [SAS-L] combine incomplete observations by var

If those are all the variables in the data set, you could do something like this (untested):

proc summary nway missing data=in_dsn; class date; var p1 p2 p3 p4; output out=out_dsn (drop=_type_ _freq_) sum=; run;

You could use PROC REPORT instead if you also want a printout at the same time.

-- JackHamilton@FirstHealth.com Manager, Technical Development Metrics Department, First Health West Sacramento, California USA

>>> "Michelle Jellinghaus" <michelle@EMODE.COM> 06/18/2003 5:22 PM >>> I have a data set that I would like to collapse based on date.

It looks something like this right now:

date p1 p2 p3 p4 4/5/03 40 . . . 4/5/03 . 30 . . 4/5/03 . . 20 . 4/5/03 . . . 10 4/6/03 50 . . . 4/6/03 . 40 . . 4/6/03 . . 40 . 4/6/03 . . . 30

And I'd like it to look something like this:

date p1 p2 p3 p4 4/5/03 40 30 20 10 4/6/03 50 40 40 30

This would provide me with one observation per date, with a complete set of values for each variable, ridding the data set of all missing values.

Any ideas would be very much appreciated.

Thank you,

Michelle