Date:         Mon, 19 May 2008 17:27:32 -0500
Reply-To:     Mary <mlhoward@avalon.net>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Mary <mlhoward@AVALON.NET>
Subject:      Re: Variance function in IML
Comments: To: Cristian Gugiu <crisgugiu@yahoo.com>
Content-Type: text/plain; charset="iso-8859-1"

Cristian,

Here's my condensed version; I can do that in one line :-)

proc iml;

start mod1;

x={10 20 30 40 50};

std=sqrt(sum(((x-j(1,ncol(x),x[:]))##2))/(ncol(x)-1));

print std; 

finish;

run mod1;

quit;

run; 



-Mary

  ----- Original Message ----- 
  From: Cristian Gugiu 
  To: Mary ; SAS-L@LISTSERV.UGA.EDU 
  Sent: Monday, May 19, 2008 4:18 PM
  Subject: Re: Variance function in IML


  Per the recommendation of one of the members, I calculated the variance from the sums of squares, as follows

  N=nrow(A);

  ss=A[##];

  sum=A[+];

  Var=(ss-(sum**2)/N)/(N-1);

  Sd=sqrt(Var);




  ----- Original Message ----
  From: Mary <mlhoward@avalon.net>
  To: P. Cristian Gugiu <crisgugiu@YAHOO.COM>; SAS-L@LISTSERV.UGA.EDU
  Sent: Monday, May 19, 2008 5:13:36 PM
  Subject: Re: Variance function in IML


  Christian,

  I didn't find where you could do these a variance or standard deviation directly on a matrix (though it seems that you could apply the formula to a matrix with other functions).  I did find that you could do it from data sets, and then use the opt{save} option in the summary to save it; the name of the matrix with the results is the variable name. 

  data test;

  infile cards;

  input x;

  cards;

  10

  20

  30

  40

  50

  run;

  proc iml;

  start mod1;

  use test;

  summary var {x} stat{std} opt{save};

  print x;

  finish;

  run mod1;

  run; 




  Producing Summary Statistics 
  Summary statistics on the numeric variables of a SAS data set can be obtained with the SUMMARY statement. These statistics can be based on subgroups of the data by using the CLASS clause in the SUMMARY statement. The SAVE option in the OPT clause enables you to save the computed statistics in matrices for later perusal. For example, consider the following statement. 


 
   > summary var {height weight} class {sex} stat{mean std} opt{save}; 
 
               SEX       Nobs  Variable        MEAN         STD 
               ------------------------------------------------ 
               F            9  HEIGHT      60.58889     5.01833 
                               WEIGHT      90.11111    19.38391 
 
               M            9  HEIGHT      64.45556     4.90742 
                               WEIGHT     110.00000    23.84717 
 
               All         18  HEIGHT      62.52222     5.20978 
                               WEIGHT     100.05556    23.43382 
               ------------------------------------------------
This summary statement gives the mean and standard deviation of the variables HEIGHT and WEIGHT for the two subgroups (male and female) of the data set CLASS. Since the SAVE option is set, the statistics of the variables are stored in matrices under the name of the corresponding variables, with each column corresponding to a statistic requested and each row corresponding to a subgroup. Two other vectors, SEX and _NOBS_, are created. The vector SEX contains the two distinct values of the class variable SEX used in forming the two subgroups. The vector _NOBS_ has the number of observations in each subgroup. 
  Note that the combined means and standard deviations of the two subgroups are displayed but are not saved. 

  More than one class variable can be used, in which case a subgroup is defined by the combination of the values of the class variables. 



  -Mary

    ----- Original Message ----- 
    From: P. Cristian Gugiu 
    To: SAS-L@LISTSERV.UGA.EDU 
    Sent: Monday, May 19, 2008 3:26 PM
    Subject: Variance function in IML


    Is there a variance function in IML? I have a matrix A. I am trying to
    calculate a 99% CI for the mean value. I know how to get the mean, A[:].
    How can I calculate the variance or standard deviation of A? I know I can
    save the matrix as a dataset and calculate these summary stats the old-
    fashioned way. However, I would prefer to do everything in Iml.

    Thanks,

    Cristian