About tradepeter

tradepeter · ‎01-08-2013

Hello all: I wonder if you can calculate percentage of nonmissing values using proc tabulate. What I need is sth. like Mean var Std N Percent of nonmissing values Although the statistic can be calculated as simple as N/(N + NMISS), I cannot seem to find a way to do that directly. Thanks, Peter

tradepeter · ‎08-13-2012

Hello all: I am trying to create a format dataset from proc format using cntlout. I use SAS version 8. This is my code: proc format cntlout=myformat; value patientgroup_f 0 = 'Case' 1 = 'Control'; run; However, this is the warning I get NOTE: The format name 'PATIENTGROUP_F' exceeds 8 characters. Only the first 8 characters will be used. It appears that the format name is too long so SAS truncate it to 8 characters. Is there any way to disable this behavior so that the full format name will be stored. Thanks, Peter

tradepeter · ‎07-17-2012

Thank you all for the help. I guess I will need a way to uniquely name my temporary datasets.

tradepeter · ‎07-16-2012

Hello all: I am wondering whether there is a way to define local/global SAS dataset? I am writing a Macro within which several temporary datasets will be created and merged. I want those temporary datasets to be local within the Macro definition so that I wouldn't accidentally overwrite any existing global datasets with the same name. I know SAS distinguishes local/global Macro variable but I cannot find an analogy for datasets. Is there any way to do that in SAS, like local/global objects in other languages such as C++ or Java? Thanks, Peter

tradepeter · ‎04-30-2012

I guess I am looking for a more general solution to display rows by their unformatted values, even when different values may have the same formatted output. For exmaple, I may need a table like Variables Options Gender Male Female Unknown Ethnic White/Caucasian Black Asian Latin American Unknown Unknown should always be the last row for each category. Apparently you cannot display by formatted value. What I did is to assign values to variables and options. e.g proc format; value variable_f 1 ='Gender' 2='Ethnic'; value options_f 1.1='Male' 1.2='Female' 1.3='Unknown' 2.1='White/Caucasian' ... 2.5 = 'Unknown'; The problem I run into is since both 1.3 and 2.5 have the same label 'Unknown'. The second Unknown is always displayed on top within the second category, since it has been displayed once before. e.g Gender Male Female Unknown Ethnic Unknown White/Caucasian Black Asian Latin American This seems to be true even when I set order=unformatted/data in proc tabulate or order=internal in proc report. The SAS document reads 'PROC TABULATE maintains for subsequent categories the order that was established by earlier categories.' http://support.sas.com/documentation/cdl/en/proc/61895/HTML/default/viewer.htm#a000146763.htm#a002473750 This is very annoying. All I want is to order the data by value and replace values by their format, even when 2 values have the same format. Any suggestions? Thanks, Peter

tradepeter · ‎04-23-2012

Hello all: I need to display statistics of several categorical variable in the same table. something like case control Age <64 5 10 unknown 1 1 Gender female 3 6 unknown 2 2 The problem I run to is for each variable, there will be a line of unknown and it has to be the last line of that category. However, SAS has some strange way to order the data. For example, the code at the end will generate sth like Age <64 unknown Gender unknown female The different 'unknown' format has different numerical values that should be sorted correctly. I also explicitly ask to order the table by value or 'unformatted'. SAS still seems to order the display according to format (Otherwise, the unknown for gender with value 2.1 should be displayed below female which equals 2.0). Is there any way to solve this? Why SAS orders tabulate in such a weired way? Thanks, data test1; input vname options group count; datalines; 1 1.0 1 15 1 1.0 2 12 1 1.1 1 2 1 1.1 2 25 2 2.0 1 10 2 2.0 2 11 2 2.1 1 23 2 2.1 2 20 ; proc format; value vname_f 1='age' 2='gender'; value noptions_f 1.0='<65' 1.1='unknown' 2.0='Female' 2.1='unknown'; value group_f 1='case' 2='control'; run; proc tabulate data=test1 order=unformatted; var count; class vname options group ; table vname*options, group * count; format vname vname_f. options noptions_f. group group_f.; run;

tradepeter · ‎04-16-2012

I tried to use CLASSDATA and EXCLUSIVE. However, tabulate will not generate correct summaries if missing value is not set as a level in CLASSDATA. If missing value is set as a valid level, the row&column for missing will still be display in tabulate. Maybe I am not doing it right.......? Thanks,

tradepeter · ‎04-16-2012

Hi all: I am wondering how to avoid displaying missing values in a table by proc tabulate. Say, for example, I have a clinical dataset with several variables such as treatment: group = 'case'/'control' frailty scale: fscale= 1,2,3 symtom: sym = 1,2,3; proc data=mydata out=test; class group; class fscale sym/ missing; table fscale sym, (group ALL)*(N PCTN) / missingtext='0'; run; I put 'missing' in the 'class' statement in order to get correct statistics. However, I don't want summary for missing values e.g Case Control fscale N PCTN N PCTN . 10 10 10 10 /* I don't want this line */ 1 0 0 5 5 ...... Is there any way to do that directly? I cannot find this option online. I tried to output the result to a table (test in this case) and remove the record corresponding to the first line. However, there is problem with this approach. Notice there is no value for Case and fscale = 1, which I printed as 0. SAS will not output it in the output table test. group fscale N... Control 1 10 /* this record is there in test*/ Case 1 0 ... /* this will not be outputed, but I need this to reconstruct my table */ Thanks,

tradepeter · ‎04-13-2012

Hi Tom: Thank you for the reply.Your example works. Just two questions, 1. Do you mean 'ods listing close;' instead of 'ods listing close;' 2. The variable values (gender in your example) and associated "F-" values from proc freq are always the same. I was thinking 'F-' variable contains the formatted value and the one without "F" contains whatever values in the table. What exactly F-variables are supposed to equal? Thanks, Peter

tradepeter · ‎04-11-2012

Hello all: In a project, I hope to create one summary table for multiple variables. For example, assume the dataset contains information about patients' gender, agegroup and disease: id gender agegroup disease 1 0 1 3 ...... The coding would be 0-male, 1-female and 0-children, 1-adults etc al. I can easily produce a summary table for each one of those variables seperately. But what I need is to stack all of them together in the same table like variable value N percentage gender male 50 50% female 50 50% agegroup children 32 32% adult 68 68% ...... However, I cannot come up with an easy solution with any of proc report, tabulate or freq. I could run multiple proc report/freq, output summaries to individual datasets and merge them together. But the code will be very long for many variables. Also, procs always output original values instead of the format (e.g output 0 for gender instead of male for gender) so I cannot merge them first and apply format later. Any help would be greatly appreciated. Thanks,

tradepeter · ‎04-10-2012

Hello all: I have a data file that contains survival times of mice under different groups. mouse.dat: "index" "Time" "Cause" "Group" "1" 159 "lymphopma" "germfree" "2" 189 "sarcoma" "control" ..... Say, I want to display the number and the percentages of mice that died due to different dieases(cause) using proc report. e.g something like N percentage lymphopma 51 0.51 sarcoma 49 0.49 However, I don't seem to be able to do that directly. /*I thought this is the correct syntax but it will generate an error*/ proc report data=mouse nowindows headline; column cause, (N PCTN); define cause / group; ERROR: A GROUP variable appears above or below other report items. /* this will run but PCTN always gives me 0*/ proc report data=mouse nowindows headline; column cause (N PCTN); define cause / group; Anyone can help with this? I hope to know why the above code does not work and how to do it correctly. Many thanks, Peter

tradepeter · ‎04-04-2012

Hello all: I want to read some characters into SAS from a text file. One record has a value '. ' which is a period followed by a space. However, SAS read treat it as missing. My understanding is that period is treated as missing for numeric but not for character. So why this is the case and how to read a value like that as it is? For example, if this is my input file mydoc.txt: a . b this is my SAS: data mydoc; infile 'mydoc.txt'; input val$; proc print data=mydoc; run; The result will be: obs val 1 a 2 3 b Any suggestions? Thanks, Peter

Online Status	Offline
Date Last Visited	‎09-01-2015 07:11 AM

percent of nonmissing in tabulate

cntlout in SAS v8

Re: local SAS dataset within a macro

local SAS dataset within a macro

Re: variables with the same format

variables with the same format

Re: not to display missing value in a table

not to display missing value in a table

Re: stack statistics of multiple variables in the same table

stack statistics of multiple variables in the same table