How to condense multiple observations
in [SAS]
|
Prev: how to use variable value as variable in the same dataset
Next: Logit-poisson; Interpreation of random effect and model diagnostics
From: Jinto83 on 13 Mar 2010 05:05 Hi, Suppose I have a file like this: Brand ReporterID SALE1998Q3 SALE1998Q4 SALE1999Q1 SALE1999Q2 Ford 2 200 300 400 Ford 1 200 500 Ford 3 300 Toyota 3 200 340 700 Toyota 2 100 250 How do I condense multiple brand information from different reporter sources into one observations. But Add a conflit_flag if there is disagreements among the diffrent reporters. So the desired output file is: Brand SALE1998Q3 SALE1998Q4 SALE1999Q1 SALE1999Q2 Conflit_in_reports Ford 200 300 400 500 No Toyota 100 200 340 700 Yes --- news://freenews.netfront.net/ - complaints: news(a)netfront.net ---
From: Tom Abernathy on 13 Mar 2010 12:14 You could use PROC MEANS/SUMMARY. If there is no disagreement then the STD will be 0. If there are no missings then N will equal _FREQ_. On Mar 13, 5:05 am, "Jinto83" <jint...(a)sina.com> wrote: > Hi, > > Suppose I have a file like this: > > Brand ReporterID SALE1998Q3 SALE1998Q4 SALE1999Q1 SALE1999Q2 > Ford 2 200 300 400 > Ford 1 200 500 > Ford 3 300 > Toyota 3 200 340 700 > Toyota 2 100 250 > > How do I condense multiple brand information from different reporter sources into one observations. But Add a conflit_flag if there is disagreements among the diffrent reporters. > > So the desired output file is: > Brand SALE1998Q3 SALE1998Q4 SALE1999Q1 SALE1999Q2 Conflit_in_reports > Ford 200 300 400 500 No > Toyota 100 200 340 700 Yes > > --- news://freenews.netfront.net/ - complaints: n...(a)netfront.net ---
From: Jinto83 on 22 Mar 2010 12:29 Thank you for your reply. But How do I 'imprint', so to speak, the results of proc means into by group. So that whenever there is duplication in at least one of the SALE variables the flag variable would be set to 1? "Tom Abernathy" <tom.abernathy(a)gmail.com> ???? news:40a6d75c-8bb7-4fe5-abbd-665c03c57e48(a)y17g2000yqd.googlegroups.com... You could use PROC MEANS/SUMMARY. If there is no disagreement then the STD will be 0. If there are no missings then N will equal _FREQ_. On Mar 13, 5:05 am, "Jinto83" <jint...(a)sina.com> wrote: > Hi, > > Suppose I have a file like this: > > Brand ReporterID SALE1998Q3 SALE1998Q4 SALE1999Q1 SALE1999Q2 > Ford 2 200 300 400 > Ford 1 200 500 > Ford 3 300 > Toyota 3 200 340 700 > Toyota 2 100 250 > > How do I condense multiple brand information from different reporter sources into one observations. But Add a conflit_flag if there is disagreements among the diffrent reporters. > > So the desired output file is: > Brand SALE1998Q3 SALE1998Q4 SALE1999Q1 SALE1999Q2 Conflit_in_reports > Ford 200 300 400 500 No > Toyota 100 200 340 700 Yes > > --- news://freenews.netfront.net/ - complaints: n...(a)netfront.net --- --- news://freenews.netfront.net/ - complaints: news(a)netfront.net ---
From: Tom Abernathy on 22 Mar 2010 16:47
On Mar 22, 12:29 pm, "Jinto83" <jint...(a)sina.com> wrote: > Thank you for your reply. > But How do I 'imprint', so to speak, the results of proc means into by group. So that whenever there is duplication in at least one of the SALE variables the flag variable would be set to 1? > > "Tom Abernathy" <tom.aberna...(a)gmail.com> ????news:40a6d75c-8bb7-4fe5-abbd-665c03c57e48(a)y17g2000yqd.googlegroups.com... > You could use PROC MEANS/SUMMARY. If there is no disagreement then > the STD will be 0. If there are no missings then N will equal _FREQ_. > > On Mar 13, 5:05 am, "Jinto83" <jint...(a)sina.com> wrote: > > > > > > > Hi, > > > Suppose I have a file like this: > > > Brand ReporterID SALE1998Q3 SALE1998Q4 SALE1999Q1 SALE1999Q2 > > Ford 2 200 300 400 > > Ford 1 200 500 > > Ford 3 300 > > Toyota 3 200 340 700 > > Toyota 2 100 250 > > > How do I condense multiple brand information from different reporter sources into one observations. But Add a conflit_flag if there is disagreements among the diffrent reporters. > > > So the desired output file is: > > Brand SALE1998Q3 SALE1998Q4 SALE1999Q1 SALE1999Q2 Conflit_in_reports > > Ford 200 300 400 500 No > > Toyota 100 200 340 700 Yes > > > --- news://freenews.netfront.net/ - complaints: n...(a)netfront.net --- > > --- news://freenews.netfront.net/ - complaints: n...(a)netfront.net ---- Hide quoted text - > > - Show quoted text - proc summary data=sales nway ; class brand; var sale: ; output out=summary ; run; data flags; set summary ; where _stat_='STD'; conflict = max(of sale:) > 0 ; keep brand conflict ; run; |