I am trying to vectorize the code below. Can somebody help me to avoid the two do loops? You can ignore the code before the do loops, that is simply to generate the data. proc iml; /*Generate Data*/ data=J(5,9,0); call randseed(1); call randgen(data, "Normal"); dim=nrow(data); obs=ncol(data); weights=J(obs,obs,0); call randgen(weights, "Normal"); mat=j(dim,dim,0); /*I would like to vectorize this loop*/ do j=1 to obs;   do i=1 to obs;   x=data[,i]-data[,j];   mat=mat+weights[i,j]*x*t(x);   end;  end; quit; Message was edited by: bernhard  I (rather Rick Wicklin actually) noticed an error in the code. In fact I need to go through the full inner loop up to obs. ... View more

Hello All From the following thread https://communities.sas.com/message/140502 I understand that under Windows (in my case Win7 Enterprise 64-bit) the maximum size of a single matrix in SAS IML is 2GB. I can initialize several 268435448x1 matrices without problems (it seems there is a need for 64 special bits since 268435448=2*1024^3-64). At the end of the thread it is mentioned that SAS is trying to increase that limit in future releases. I was wondering if there are any news about that? What is the limit in SAS 9.4 and IML 13.1? I know that I can probably alter my algorithm(s) to work with smaller matrices (probably by thinking a bit about it and using smarter programming or possibly via splitting up the data and iterating), still I am curious and would appreciate if that limit could be increased. ... View more

Hello I am trying to find the equivalent of Excel's large function which returns the k'th largest element in a list. My problem is the following: "Given a dataset with, say, 10^6 observations and 25 variables return the kth largest element for variables x1 to x25." I tried a few things with proc sort, proc rank, data steps (obs=, firstobs=), _N_, some sql code. I also tried percentiles, but none of the percentiles options (there are 5 different ones) do exactly correspond to the kth largest element. Thus, none of the solutions I tried are really satisfying. The main issues I have are: 1. I do not want to execute some code (say a macro) for each of the variables. Let's say I do not want to sort my data set 25 times. This is to slow. 2. Proc Rank seems to work reasonably fast and it can work on all of the variables at the same time. However, none of the TIES= options suits my needs, since I want the ranks to be 1,2,3,4,..., 999999, 1000000 (simply 1-n all integers without interleaving any numbers) Can anybody help me with that? I timed my code again. It turns out that sorting 25 times (for all different variables) takes about 9 seconds (using "sasfile ds open"). I think I can actually live with that. Even though proc rank can give me all the ranks in 2.5 seconds. Bernhard ... View more

Hello all My questions applies to the min and max function. In the following I will use the min function for simplification. The help function on min says „The argument list can consist of a variable list, which is preceded by OF.” I was wondering whether it is mandatory to use the “of”. In fact I have been using the min function for a long time never being aware of this. So far I am neither aware that leaving out the min caused problems. Recently a colleague of mine pointed out to me that he has seen examples where this can cause problems (however, he cannot recall the specific case). Thus he suggested to always use “OF”. Now this question came up since we have quite some programs that use min in dozens of places and we want to find out whether we should amend the programs or not. In fact this leads to a few follow up questions. Usually I make use of variable operations as well as the usage of macro variables in min. To give you an example: %let mvar=2; data  _null_; a=9; b=1; c=min(a,b,a*&mvar.,a*b+b,a*99,b/&mvar.+1); put c; run; So my question is whether using min in this way will always yield the result I had in mind (which is canonical). It seems to me that if I was to meticulously apply the min function as described in the help function, I would need to define additional variables. For example already a simple statement such as “c=min(of a*b,b)” results in an error. Has anyone encountered unexpected behaviour with the min function by not using the "of" (or because of other reasons) like my colleague has? Regards Bernhard ... View more

Hello all I am using SAS for a while now but still have not figured this out: Currently I use SAS 9.3 on Windows 7 and what I would like is that the explorer view is set to details (without tree view) when I start SAS. If possible I would also like to have the output window closed. Right now I always have to click View->Details. Is there a way to achieve this via the settings or maybe a tiny macro at startup? Admittedly this is not a serious problem at all, but one can still ask about it. Regards Bernhard ... View more