Statistical programming, matrix languages, and more

Conversion of matrices in a dataset

Reply
Frequent Contributor
Posts: 84

Conversion of matrices in a dataset

Hi

I am working in SAS/IML

I have generated multiple matrices, each of single row and two column using do loop

now I want to convert all the matrices in a single data set with only variables and number of observation would definitely be equal to number of iterations

Kindly help me

following is the part of matrices output

i                       I_ij               

  1            -17.98802     11.706846

                          

   i                       I_ij               

2            -16.85767    12.894326

                          

  i                         I_ij             

3            -18.33006    11.883216 

Frequent Contributor
Posts: 143

Re: Conversion of matrices in a dataset

If you are wanting to create a SAS dataset with the results, then look in to the CREATE and APPEND statements.  For example:

     I_ij = j(1,2);
   create mydata from I_ij;

   do i=1 to 10;
     call randgen(I_ij, 'uniform');
     append from I_ij;
   end;

   close mydata;

Frequent Contributor
Posts: 84

Re: Conversion of matrices in a dataset

Thank you very much

in your example, you create a matrix then draw sample matrices through uniform dist. and then create data set for all drawn sample

but in my case i dont want to draw samples, however i have multiples matrices now my objective is to create a data set involving all matrices,

for this purpose,during my search, I found following example, which is not exactly what I want, but may be useful

Proc iml;

PCOR = {1.00 0.40 0.40,

0.40 1.00 0.40,

0.40 0.40 1.00};

dsNames = {A B C};

print dsNames;

do i = 1 to ncol(dsNames);

pvar = diag(vecdiag(pcor))*i;

create(dsNames) from pvar;

append from pvar;

close (dsNames);

end; */;

The problem with example is that when I run this there is an error in  create(dsNames)   and  close (dsNames)

Pls guide

Thanks

Frequent Contributor
Posts: 143

Re: Conversion of matrices in a dataset

The code you posted works for me.  I am running SAS 9.4 (IML 13.1) and it is possible you have an old version without this feature.

I think there may be a way around with call execute.  Try replacing the create and close statements with something like:

call execute('create ' + dsNames + ' from pvar;');

call execute('close ' + dsNames + ';' );

This will only work if the code is inside a module.

Frequent Contributor
Posts: 84

Re: Conversion of matrices in a dataset

Thanks

yes i am using sas 9.02 version

I think version is matter here

can u share me sas 9.04 software backup through dropbox

I would be really grateful to you

Regards

Frequent Contributor
Posts: 84

Re: Conversion of matrices in a dataset

Dear  Ian Wakeling

I would really be grateful to you if you could share me SAS 9.4 version backup

Regards

Frequent Contributor
Posts: 143

Re: Conversion of matrices in a dataset

No, I can't do that.   You need to talk to someone at your nearest SAS office about an upgrade.  Or perhaps investigate the SAS University Edition.

Frequent Contributor
Posts: 84

Re: Conversion of matrices in a dataset

I got SAS 9.3

Now try to run following program

Proc iml;
PCOR = {1.00 0.40 0.40,
0.40 1.00 0.40,
0.40 0.40 1.00};

dsNames = {A B C}; /* specify names of data sets */
do i = 1 to ncol(dsNames);
pvar = diag(vecdiag(pcor))*i;
create(dsNames) from pvar; /* create work.A, then work.B, and so on */
append from pvar;
close (dsNames);
end;

But it is not working gives few errors

Log is as follows

NOTE: IML Ready

5088  PCOR = {1.00 0.40 0.40,

5089  0.40 1.00 0.40,

5090  0.40 0.40 1.00};

5091  dsNames = {A B C};

5091!                    /* specify names of data sets */

5092  do i = 1 to ncol(dsNames);

5093  pvar = diag(vecdiag(pcor))*i;

5094  create(dsNames) from pvar;

            -

            22

            76

ERROR 22-322: Expecting a name.

ERROR 76-322: Syntax error, statement will be ignored.

5095  append from pvar;

5096  close (dsNames);

            -

            22

            200

ERROR 22-322: Syntax error, expecting one of the following: a name, ;.

ERROR 200-322: The symbol is not recognized and will be ignored.

5097  end;

ERROR: END does not occur within DO group at line=5097 col=1.

Kindly help me

Frequent Contributor
Posts: 143

Re: Conversion of matrices in a dataset

This is the same error as before, so the ability to specify the dataset name from a character matrix must be a more recent innovation. Try this instead:

Proc iml;

start main;
PCOR = {1.00 0.40 0.40,
0.40 1.00 0.40,
0.40 0.40 1.00};

dsNames = {A B C}; /* specify names of data sets */
do i = 1 to ncol(dsNames);
pvar = diag(vecdiag(pcor))*i;
call execute('create ' + dsNames + ' from pvar;');
append from pvar;
call execute('close ' + dsNames + ';' );
end;
finish;

run main;

quit;

SAS Super FREQ
Posts: 3,630

Re: Conversion of matrices in a dataset

Same answer as before: YOu need SAS 9.3m2 or later (SAS/IML 12.1): Read data sets that are specified by an array of names - The DO Loop

SAS Super FREQ
Posts: 3,630

Re: Conversion of matrices in a dataset

What do you intend to do with the output data?  In my experience, one data set with k observations is almost always easier to work with than k data sets with one observation in each.  The ideas in this blog post might be relevant: Simulation in SAS: The slow way or the BY way

Ask a Question
Discussion stats
  • 10 replies
  • 600 views
  • 0 likes
  • 3 in conversation