If you are looking at the file using Excel, or other spreadsheet program, I recommend looking at it using a text editor such as the SAS editor or Wordpad, or a viewer such as SAS FSLIST. If you are seeing commas in the fields the quotes that are in the file are ignored by most of the programs that open CSV by default. ... View more

Or if you are reading from an external text file or data lines use the INFORMAT YYMMN6. data test; informat date yymmn6.; input date; format date mmddyy10.; datalines; 200405 200505 200507 200610 ; run; proc print; run; ... View more

I think you'll find that the FREQ option truncates fractional values. Re-run your example without the decimal parts of NUM and you'll get the same result. This calculates the correct percentages but you'll probably not be happy with the vertical axis. I find most of the times when I want to use weighted statistics that I have to pre-summarize the data prior to display. proc gchart data=test1; vbar year / discrete subgroup=cat sumvar=num raxis=none inside=subpct width=20; run; quit; Message was edited by: ballardw Message was edited by: ballardw ... View more

Hint: If you want to have the format available all the time put it in a permanent library by adding the LIBRARY= option to the PROC step. Reference e that library in the format search path using the FMTSEARCH system option or other tool. ... View more

> Hi all, I'm currently learning Basic SAS and am > trying to learn the mechanics of DO Loops. I have a > simple code here but would like to understand how SAS > would execute it. Here is the code: > > data balance; > bal=1000; > do i=1 to 12; > bal=bal*1.01; > output; > end; > > Although I do know what the result would be, I still > have a few questions that I would like to ask that > would better help me understand how SAS executes this > loop. > > 1. assume SAS is reading this for the first time and > is on its first loop (i=1). Since we're given that > bal=1000 outside the DO Loop, how come IN the Do > Loop, SAS will calculate it as bal = 1000*1.01 and > not 1000 = 1000*10.1? Shouldn't the variable be > different say total=bal*1.01? > Things on the LEFT side of a = sign are where results go (generally, there are exceptions but don't apply to this example). So BAL= in effect says "do what's on the right side of the equal sign and store the result in a place named BAL. If BAL already has a value it will be replaced. If you don't want to replace the value then don't. In your example of TOTAL=BAL*10.1; however you would get the exact same value for TOTAL. My take on the original code is someone wants to look at something like compound interest applying 1% successively. Try this code to see what's happenng. PROC Print data=balance; run; > 2. so after SAS goes through the loop once, the > balance after 1000 is multiplied by 1.01 and gives > 1010, so how does SAS memorize the 1010 for the 2nd > loop execution and multiply that by 1.01? Does SAS > change the bal=1000 outside the DO Loop to bal=1010? > If not, then how is SAS able to memorize the result > from the previous loop and carry it into the next > loop for calculation? > > Hope my questions are not confusing...thanks! Conceptually BAL is a single storage bucket that only ever contains one value at a time within a single data step. ... View more

Try this: %macro gen_vars (day=, mon=, mon_num=, year=, year_num=); %local i j result; %do i=1 %to &year_num %by 1; /*Loop over years*/ %do j=1 %to &mon_num %by 1; /*Loop over months*/ /* here's the part you were missing to get a single output space delimited list*/ %let result=&result _&day._%scan(&mon, &j)_%scan(&year, &i); %end; %end; &result %mend gen_vars; %let mon=Jan Mar Jun Sep Dec; %let year=95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11; %let varlist = %gen_vars(day=1, mon=&mon, mon_num=4, year=&year, year_num=17); %put &varlist; Your sytem settings for maximum macro value size may have an effect if this generates a long enough list. If you nee ... View more

Or without knowing how many levels of match/mismatch are needed, either fixed such as first two as shown, running two where the second of your displayed pairs could then be paired with a third, or an indefinate number. ... View more

If you are using F8 or other key to submit your job, you could change the entry in the KEYS menu item to: Log;clear;submit;log; This will clear the log before each program submitted. ... View more

> ballardw, > > I tried, but this only gave me the number of levels > each variable has. In fact, what i want to find out > is the names of all the levels, not just 'how many > levels'. Thank you. Use the list of varialbles that look likely to be categories in Proc Freq. You could just use Proc Freq data=yourdata;run; BUT if you have identification variables (which are categorical), phone numbers and such as well as any continuous or pseudocontinuous (income for examp) but don't want to see every one of those you need a way to find the likely ones of interest. That's what the levels statement does in this case, give a starting point. I'd be tempted to dump the levels output to a dataset and filter on the range of levels to get a list of variable for use elsewhere. ... View more

> Hi All, > > Thanks for the responses. It appears I confused > people by asking about "incrementally" calculating > the median. The idea was to avoid the use of PROCs > since I need to generate metrics that either would > require multiple SAS PROCs or DATA Step code. > Instead of doing 3 or more passes on the data, I'd > like to try to do it in a single pass using one DATA > Step. > > I can certainly hold all (or perhaps 1/2) of the > unique values in memory and do a calculation on those > values at the end of the DATA Step but was wondering > if there was another option. > > Thanks, > > Tim Stearn By any chance did you mean: generate metrics that either would NOT require multiple SAS PROCs or DATA Step code As mentioned above, a single pass through PROC MEANS or SUMMARY would probably be the most efficient both in time of code development and execution time. ... View more

The quick simple view would be PROC FREQ. If you leave out a table statement you'll get output for all variables in the data. Use the NLEVELS option as Proc freq data= nlevels;run; This generates a table of variable names, labels and how many levels (categories) a variable may have. Large numbers would indicate things that aren't likely to be "categories" in the way you are thinking. ... View more

If the data are more complicated then I understand. It may be that the better idea would become to create additional "rate" variables if you have the confidence limits for the X values. (Pseudo code). rate=trate;ratetype="T";output; rate=trateuppercl;ratetype='TCU';output; rate=tratelowercl;ratetype='TCL';output; rate=prate;ratetype="P";output; rate=prateuppercl;ratetype='PCU';output; rate=pratelowercl;ratetype='PCL';output; And appropriate symbol statements to join the limits. ... View more

Have you tried the I=R type symbols? I think 2 symbol statments with i=RLCLI95 or i=RLCLM95 depening on whether you're looking for limits of individual or mean estimators. The online help for Gplot should have an example called Plotting Two Variables which does a regression and confidence limits around one set of variable pairs. Two separate symbol statements with different colors and symbols should overlay. One big advantage is you just drop the data in and let the regression be handled by Gplot. If you're confidence limits come from some thing other than linear, quadratic or cubic regression, then I'm out of ideas. I think a properly constructed annotate data set shoulc still work but I haven't time to figure out how to display it at the moment. ... View more

It might help to show what you've attempted. ... View more

The horizontal axis need to use the same X variable. Your code uses two, TRATE and PRATE. One approach to getting what I believe you are attempting is to create a single X variable and a categorical variable to differentiate. Quick and dirty example: data newtest (keep=yvar rate ratetype); set test; rate=trate;ratetype="T";output; rate=prate;ratetype="P";output; run; proc gplot data=newtest; plot yvar*rate=ratetype /haxis=axis2 vaxis=axis1; run; quit; Since you don't have any details on the annotate I'm not including it. ... View more