Or for use in a where clause Where not ((weekday(datepart(date))=6 and timepart(date) ge "23:00"t) or weekday(datepart(date))=7 or (weekday(datepart(date))=1 and timepart(date) le "23:30"t)); ... View more

Use multiple class statements. You can have as many as you like: Class var1 / order=data; Class var2 / order=descending; Class var3 / order=formatted; ... View more

proc sql; create table test3 as select * from test1 outer union corr select b.* from (select distinct x from test1) as a left join test2 as b on a.x=b.x order by x; quit; May execute faster on larger datasets. ... View more

I would generally use: If upcase(BodyStyle) in ('3D HATCHBACK','4D SEDAN',...,'2D SOFTTOP') because so many data sources tend to mess with case. The other thing to look for would be different numbers of spaces between your list in the "IN" clause and your data. ... View more

Maybe not the most elegant solution but I have made similar tables by concatenating the two (or more) variables in another data set. Data tab; set query; /* probably need some LENGTH and other character stuff to make this pretty*/ metier_emploi = metier || emploienclair; run; Proc tabulat data=tab; class metier_emploi Niveau; TABLE metier_emploi , Niveau*n=''; run; ... View more

> I would like to be able to name the dataset for each > distribution in turn without having to change the > value of the &outdata macro variable each time. I > would like it to be able to refer to the > &distribution macro variable rather than having to > change both macro variables for each distribution in > turn. If that makes sense? > > Message was edited by: den Time to give a better example of what you are looking to do. It sounds like you are attempting to build multiple datasets with one pass but you only have one assignment to DISTRIBUTION at a time, so only one output set would be created. Did you run my example code and did it create a data set named with the distribution or not? Or are you wanting the parameters involved in the data set name as well? That may require things not available in open code. ... View more

> Dear all, > > I was wondering if anybody knows the answer to the > following? I have a macro listing various different > statistical distributions. If I wish to simulate > values from a certain distribution, I change the name > of the macro variable which refers to that > distribution. > > eg %let distribution='Normal' > ...... > if &distribution="Normal" then > distribution=rand('Normal',0,1); > > I would like to be able to call output generated by > the distribution by the corresponding distribution > name. For the above example the output dataset name > would be "Normal". I have tried writing a new macro > variable eg %let outdata=&distribution but this > doesn't appear to work. Any thoughts on how to > resolve this would be much appreciated. I think you main issue is defining the macro variable with the quotes. Try: %let distribution=NORMAL; data &distribution; if upcase("&distribution")='NORMAL' then x=rand ('NORMAL',0,1); run; ... View more

Look at documentation for FINFO function. The information about external files depends on your OS and file system. There should be an example that will let find out all of the file information the system will return. ... View more

Another thing is that RETAIN can be used to set an intial value such as 0 for the variables. ... View more

A tad tongue it cheek, but if the IP is worth enough it may be feasible to put the code on a thumb drive or similar media, visit the client, complile the code on their machine and not leave a copy of source. If you could tie this to another customer service reason, so much the better. ... View more

Here's a rough skeleton of what I think you're asking: Assuming you have an ID variable (patient number, health card, what ever) and that the date of admission is a SAS date value: Proc sort data=[your data set name]; by patient_id admin_date;run; /* or what ever you have named your ID and date variables*/ Data [new data set name]; set [your data set name]; by patient_id; last_admin_date=lag(admin_date); date_dif = admin_date - last_admin_date; /* create some 0/1 coded variables to get counts by summing*/ if first.patient_id then do; readmin_7 = 0; readmin_14 = 0; readmin_21 = 0; readmin_30 = 0; readmin_more= 0; date_dif = 0;/* 0 days since last visit for first*/ end; else do; readmin_7 = (date_dif le 7); readmin_14 = (date_dif le 14); readmin_21 = (date_dif le 21); readmin_30 = (date_dif le 30); readmin_more= (date_dif gt 30); end; label readmin_7 = 'Readmit within 7 days' readmin_14 = 'Readmit within 14 days' readmin_21 = 'Readmit within 21 days' readmin_30 = 'Readmit within 30 days' readmin_more= 'Readmit more than 30 days' ; format admin_date last_admin_date mmddyy10.; run; /* you could drop the last_admin_date and the date_dif variables after your comfortable they are working correctly. */ /* get a count by visit of each readmin group*/ proc tabulate data=[new data set name]; class patient_id; var readmin_7 readmin_14 readmin_21 readmin_30 readmin_more; table patient_id='Patient, (readmin_7 readmin_14 readmin_21 readmin_30 readmin_more)*sum=''; run; ... View more

If FLUPOS is coded as 0/1 then MEAN instead of PERCENT will give you a percent of 1 coded responses, use an appropriate format for display. ... View more

An example of this being useful comes with files that have mixed record content resulting in different lengths in an irregular occurence pattern. Consider a data logger that has period output at hourly and daily intervals but may contain an event triggered record at any time of day and each of these record formats has a different length. And then throw in a possibility of an incomplete record because of data transmission issues. If you read past the end of a record you can test for conditions to output diagnostics of approximately where the data file has problems and still capture the 'valid' data up to the break. I'm sure that's not the only use but the one I remember dealing with that used this. ... View more

> That might make the annotation datasets easier to > build, but how would that help render the graphics > any smoother? > > -Max Part of the issue is larger item shifting the apparent center. The SLICE would fix that part. Depending upon display, system and other factors it MAY render differently than font images, possibly reducing the not quite roundness. ... View more