Dear SAS experts   I would like to identify for the variables num1-num3 observations which have for one of these variables a value >18. The _numeric_ code in an array statement seems useful to identify num1-num3, but it also includes the variable Age, as Age is also numeric. Is there any way to identify in an array all numeric variables in a dataset except for one (specified) numeric variable?   See the example code below.   data have; input Name $ Age num1 num2 num3; datalines; Maria 30 5 2 17 John 32 3 7 20 ; run;   data have (drop=i); set have; disk=.; array missmake {*} _numeric_;                  /*Is it possible to somehow exclude Age from the array?*/ do i=1 to dim(missmake); if missmake(i)>18 then disk=1; end; run;   Thank you ... View more

Dear SAS experts   I would like to create a categorical variable that include all observations that sum up to certain values based on two numerical variables.   Given this simple dataset:   data example; input value1 value2; datalines; 2 5 3 5 6 10 2 4 3 5 2 5 ; run;   I want to categorize the observations (as they are ordered here) based on running sums. When the running sum reaches >=5 for value1 AND >=10 for value2 a category should be created. Then, the process is repeated again. The resulting dataset should look like:   value1 value2 cat_var 2         5         1 3         5         1 6         10       2 2         4         3 3         5         3 2         5         3   For category 1 it takes two observations before the conditions described above are met. For category 2 the condition is already met in the 3 rd observation and therefore only one observation is included in category 2. For category 3, while the condition is met for value1 already in the 5 th observation for value2 it is met on the 6 th observation. Therefore 3 observations are included in category 3.   Does anyone have a suggestion on which syntax I can use to create such a categorical variable?   Thank you ... View more

Dear SAS experts   I am a relatively new user, but have used Stata for some years.   I would like to be able to generate random numbers uniformily distributed. I want to do this seperately for certain subgroups in my dataset, although it might actually suffice to do it for the whole dataset.   As an example:   data example; input name $ value; datalines; Maria 2 Maria 3 Maria 6 John 2 John 6 Peter 2 ; run;   I would like to randomly select a 'Maria' observation and a 'John' observation (only 1 Peter, no selection needed) such that I am left with only one of the values of 'value'. To the first issue: I thought it would be easy to generate random data for the whole dataset:   proc sql noprint; select count(*) into :N from example; quit;   data example (drop=i); set example; do i=1 to &N; uni_ran=rand("Uniform"); output; end; run;   But the amount of observations in the dataset goes from 6 to 36 - why?   The second issue that I would like to be able to generate random numbers seperately for 'Maria' and 'John'. In general, how would one go about doing this? I have tried by including a -by- but I cannot get it to work. I imagine I have to somehow confine the do loop to the span of observations in each category, e.g. 3 for Maria and 2 for John, perhaps using some functions which identifies the amount of observations in a category of a certain variable. In the actual dataset I am using there are in many cases >3 observations within each category.   In general I have an interest in finding out how to perform different operations for each subgroup of a dataset without having to qualify the subgroup using fx -if- and -where', i.e. just "do this for each level of var". If there are 1000 levels of categorical variable this would require many lines of code using -if-.   Does someone have a suggestion on how I can resolve the issues mentioned above?   Thank you   ... View more