Someone sent me four text files.  I'm using PROC IMPORT to bring them in.  Three of them worked fine but the other gives me an error.   I think it will be messy to copy and paste the whole file so maybe someone can help me figure it out just from the Log window message.  Here is what it says.   Unable to sample external file, no data in first 5 records. ERROR: Import unsuccessful. See SAS Log for details. NOTE: The SAS System stopped processing this step because of errors. NOTE: PROCEDURE IMPORT used (Total process time): real time 0.02 seconds cpu time 0.01 seconds   No data in the first 5 records?  It looks to me like there's data from the very beginning.  Any assistance is appreciated.   BTW, I've always found it odd that it says in the SAS Log "See SAS Log for details"  Also, SAS has been around so long that you'd think that they could put possible solutions in the log when an error happens.  I mean, this will probably get solved by someone here telling me "Maybe the problem is X" and then I check X.  Since there are people that know what the potential problems are for a particular error message it seems that SAS could just add to the error message in the Log window "Potential problems are X, Y and Z."  Then peope could try those on their own and if still not successful then they could come here or go to SAS Support.  Just a thought. ... View more

  I'm running PROC LIFETEST with a four-level group variable that is defined by age (<60 vs >=60) and depression (none vs mild or more).  It runs fine.  I have an ODS SELECT statement for SAS to only generate what I need and it is as follows.   ods select Means CensoredSummary HomTests SurvDiff SurvivalPlot;   Then I change the depression dichotomy to "none or mild vs moderate or more."  Everything else is the same.  And when I run it I get the following error.   ERROR: Floating Point Zero Divide. WARNING: Output 'SurvivalPlot' was not created. Make sure that the output object name, label, or path is spelled correctly. Also, verify that the appropriate procedure options are used to produce the requested output object. For example, verify that the NOPRINT option is not used. ERROR: Termination due to Floating Point Exception NOTE: The SAS System stopped processing this step because of errors. WARNING: The data set WORK.SURV may be incomplete. When this step was stopped there were 1485 observations and 7 variables. WARNING: Data set WORK.SURV was not replaced because this step was stopped.   It still runs but it does not give me the pairwise comparisons nor do it give me the K-M plot.  If I take "SurvDiff" out of the ODS SELECT statement the the pairwise comparisons aren't there (since we're no longer asking for them) but the K-M plot does come back.   I've run models on many different survival outcomes and every time it's the same story, namely the "none vs mild or more" dichotomy always gives me the pairwise comparisons and the "none or mild vs moderate or more" dichotomy never gives them to me.   The "zero divde" makes me think it's trying to divide by 0 somewhere.  In a cuople of the models it's true that I have 0 failure for one of the four gruops.  But for other models I don't have 0s and yet I still get the error.  Even if a 0 for one of the four subgroups meant that it couldn't be used in a pairwise comparison I don't know why SAS couldn't just give me the pairwise comparisons that can be made and put asterisks for the ones that can't.   Okay, so that's my problem, any insight on this is greatly appreciated. ... View more

I have cigarette smoking data along with a continuous variable of interest.  I compute how many cigs were smoked in the last 15, 30 and 60 minutes before the measure was taken.  It turns out the the number of cigs smoked in the last 15 minutes is 0, 1 or 2, in the last 30 minuts is 0, 1, 2 or 3 and in the last 60 minutes is 0, 1, 2, 3, 4 or 5.  Okay, that's all fine.   Then what I want to do is make a vertical bar chart.  On the y-axis is the mean measure of interest and on the x-axis is separate bars for the number of cigs smoked.  I'm using PROC GCHART.  When I do it for 15 minutes it works fine and vertical bars at 0, 1 and 2 show up.  When I do it for 60 minutes it works fine and vertical bars for 0, 1, 2, 3, 4 and 5 show up.  Okay, still fine.   But when I do it for 30 minutes, instead of getting vertical bars at 0, 1, 2 and 3 as I should, I get vertical bars at only 0 and 3.  (As an aside, although it works for 15 and 60 minutes, the x-axis ticks don't say 0, 1, 2, etc and instead say 0.0, 1.0, 2.0, etc.  Maybe that's a clue in all this.)   In an AXIS statement my code has the line   order = 0 to 5 by 1;   But the bars for 1 and 2 don't show up.  (Or if I change 5 to 3 then the same thing happens.)   Then I tried order = 0 to 5 by 0.1 and still the bars for 1 and 2 didn't show up.   Then I tried 0 to 5 by 0.01 and the bars for 1 and 2 DID show up.  It's all squished together because I have so many points on the x-axis so I can't tell exactly where they're at, be it 1.99 or 2.01 or whatever, but the bars at (or close to) 1 and 2 are there.   Why is this?  I want to use 0 to 5 by 1 (or maybe 0 to 3 by 1) and have bars at 0, 1, 2 and 3, but I can't seem to accomplish this, I guess because SAS somehow thinks the value aren't exactly 1 and 2.   I did a PROC FREQ with a format 30.28, to see 28 numbers to the right of the decimal for 0, 1, 2 and 3 and every single number to the right of the decimal is 0, e.g., it's 2.0000000000000000000000000000 rather than something else.   Any ideas?  When I created these numbers I set a counter to 0 and then if a person smoked I cig I incremented the counter by 1 so I don't see how the values can be anything other than exactly 0, 1, 2 or 3.   Any info is appreciated.   Signed, Mystified in Chapel Hill ... View more

  I'm using PROC SGPANEL.  In the first panel I have side by side boxplots for Males and Females for Group A.  In the next panel same thing but for Group B.  Etc.  Five groups total.   But the person I'm doing this for doesn't want the whiskers extending from the boxplot to go to the highest observation that is below the upper fence, like the default is now, and understandably so because it's hard to explain when submitting it to a journal and nobody reading medical journals even knows what the upper fence is.  Instead this person wants the whiskers coming out of the boxplot to go to the 10th and 90th percentile.  Or maybe 5th and 95th.  Or maybe 20th and 80th.  They want to try a few things and see what looks good.  Since there are a couple extreme observations, having the whiskers go to the min and max is not good because it makes the box part get all scrunched and small.   In looking into SAS Help I see the HIGHLOW statement so that may be part of the solution.  But am I going to be able to do that for 10 groups (five for each of Male and Female) and still have side by side boxplots for Male and Female in five panels?  It seems too customized.  It would be really nice if I could just use the code I'm using now except adding a line that tells SAS what percentile to take the whiskers to, but I don't think that's possible.  I'd be happy to be proven wrong on that.   Any info is greatly appreciated. ... View more