About Reeza

Reeza · ‎03-16-2011

Hi All, I'm analyzing some survey data and want to present some data in a horizontal bar chart. The data is in terms of agreement, neutral or disagreement. I can get the data in the chart but not in the order I'd like. I would like to have the Questions order by the Agree %, so the topic that most people agree with is at the top of the charts. Second I would like the 'bars' within the questions arranged from Agree to Neutral to Disagree. I could drop the Disagree without changing the appearance, but thought I'd leave it in for now. I'm using a sample for Rob Allisons page. Here's what I have so far. Thanks! [pre] data have; input question $ agreement $ percent; cards; Q1 Agree 0.25 Q1 Neutral 0.30 Q1 Disagree 0.45 Q2 Agree 0.55 Q2 Neutral 0.10 Q2 Disagree 0.35 Q3 Agree 0.45 Q3 Neutral 0.25 Q3 Disagree 0.3 ; run; proc sort data=have; by agreement descending percent; run; goptions gunit=pct htitle=4 ftitle="albany amt/bold" htext=2 ftext="albany amt"; axis1 label=none; axis2 label=none order=(0 to 1 by .1) minor=(number=1) offset=(0,0) value=(h=3pct); legend1 label=none position=(bottom) shape=bar(3,3) across=3; /* pattern v=solid color=red; */ pattern1 v=solid color=cxbd0026; /* kind of a seafoam green */ pattern2 v=solid color=cxc7e9b4; /* this is the hex rgb color for mild blue */ pattern3 v=solid color=cx43a2ca; /* reddish color */ title1 "Percent Agreement"; proc gchart data=have; hbar question / discrete type=sum sumvar=percent subgroup=agreement/* this controls the coloring */ nostats maxis=axis1 /* midpoint axis */ raxis=axis2 /* response/numeric axis */ autoref /* reflines at every major axis tickmark */ legend=legend1 coutline=same ; run; quit; [/pre]

Reeza · ‎03-16-2011

Have you set the nowait/noxsync options if you're using DDE? Also if the process is saving over a file and you say don't save or some other excel prompt comes up you may cause the session to be interrupted. I try and make sure Excel is closed, especially the file I'm working on when using DDE. HTH, Reeza

Reeza · ‎03-15-2011

I think you need to look at processing data with a macro variable LIST rather than array. See this paper: http://www2.sas.com/proceedings/sugi30/028-30.pdf HTH, Reeza

Reeza · ‎03-15-2011

I think you need to look at processing data with a macro variable LIST rather than array. See this paper: http://www2.sas.com/proceedings/sugi30/028-30.pdf HTH, Reeza

Reeza · ‎03-15-2011

what about something like this: proc sort data=have; by x y z; run; proc means data=have n nmiss; by x y z; var a; run; X Y Z do not have to be numeric. In this example A would have to be numeric. If its not numeric you can do your conversion mentioned. Or you could apply a format to it. There's a neat trick in the paper by Lois Levin (Paper 001-30). Add in the line By x y z; and replace the _numeric_ and _character_ by what type your variable is and delete the other. [pre] proc format; value $ missfmt ' '="Missing" other="Not Missing" ; value nmissfmt . ="Missing" other="Not Missing" ; run; proc freq data=home.ex2; tables _numeric_ _character_/missing; format _numeric_ nmissfmt. _character_ $missfmt.; run; [/pre]

Reeza · ‎03-03-2011

Here's a sql solution since all the other are datastep and its what you were looking for initially. I doubt its more efficient, but useful to know. [pre] data have1; input var1 $1 var2; datalines; a 1 b 2 c 3 d 4 ; run; data have2; input var1 $1 var2; datalines; b 5 c 6 ; run; proc sql; update have1 as h1 set var2=(select var2 from have2 as h2 where h1.var1=h2.var1) where var1 in (select var1 from have2); quit; [/pre]

Reeza · ‎03-03-2011

I'm still not sure that Teradata knows how to interpret the datepart function is pass-through code. See valid functions here: http://support.sas.com/resources/papers/teradata.pdf However, I am unfamiliar with that specific db so possibly post your question on SAS-L for more responses. Cheers, Reeza

Reeza · ‎03-03-2011

you have a format in there, yymmddn8. Is that a valid sas format or a custom format that's been defined?

Reeza · ‎03-03-2011

I'm assuming the query works without the where clause or when the date you're comparing to is a full date (ie not using datepart)? It would help if you posted the log with the code and error message, if possible, XXXX'ing out any confidential info.

Reeza · ‎03-03-2011

What's the error? Here's a few things you can look into: 1) Does SQL pass through the function your using? See http://support.sas.com/kb/23/194.html 2) Is the macro variable resolving properly?

Reeza · ‎02-28-2011

Use retain with by groups and a found flag to indicate if the variable has already been set. [pre] data want; set have; by id; retain found; if first.id then found=0; flag=0; if v1 ne 0 and found=0 then do; flag=time; found=1; end; drop found; run; [/pre]

Reeza · ‎02-28-2011

A column in SAS can be only character or numeric, not both. Mixed will set columns that contain both text and numbers to characters. I don't understand this statement: If first 15 rows has char data and after it is numeric, then I need to set that variable as numeric. Are the character variables incorrect then or would you still like to see them? If you still want to see them then the column must be character. If its incorrect data then import as character anyways and then clean your data in a data step to have an audit trail of whats done rather than rely on SAS to automically clean the data in the import statement. You can check the documentation here: http://support.sas.com/documentation/cdl/en/acpcref/63184/HTML/default/viewer.htm#a003103761.htm The documentation also had details on the guessing row option. Message was edited by: Reeza

Reeza · ‎02-28-2011

The GUESSINGROWS statement applies to text/delimited files only, not Excel. If using DBMS=EXCEL or EXCELCS, you can try the Mixed=Yes option. Mixed=YES will examine columns that contain both text and numbers and set them to character type. Or export to CSV and use a data step to read the file, where you have complete control.

Reeza · ‎02-24-2011

My opinion only... I use views for reports/datasets where a dataset will be updated and I want this other data set to be updated to reflect this. A view is a step that is run every time it is accessed, rather than just reading a table. The speed depends on the calculations in the view. I also have to balance the length of time the view takes to load and how often it is used. For example if a view takes 2 seconds to run but is run 100 times a day, I'd rather have a table updated. On the other hand if it takes 30sec to run, but is run ten times a month, perhaps a view is fine. Another consideration to use a view over a dataset is storage space, but this only becomes an issue if you continue to use a lot of views, then you can 're-use' space. You still need the space to complete the task in the view. Not sure I answered your question but hope it helps.

Reeza · ‎02-24-2011

Look up the scan () function. scan ('/explore/hospitality/index.cfm', 2, '/');