doing (relatively simple) matrix algebra within spss syntax

[Michael Parent]

Hi, all,
I'm struggling with an spss syntax issue. I need to perform some basic analyses on a saved covariance matrix. I can't for the life of me get this to work under normal syntax or within matrix-end matrix syntax.

I've got covariance matrices (of varying size). They're saved as spss saved data files; the only data in the files are the actual covariance matrix data (variables labelled var1, var2, …vark, in the spss variable view but NOT as a column in the data set). For the purposes of what I need to get, I don't actually need the N.

ALL I need to do is get the sum of variances (i.e., diagonal elements) and sum of covariances (i.e., one set of the symmetrical half-diagonal elements) from the data sets (I have a few data sets but doing them one at a time is fine). I can't get it to work at all trying syntax commands or using matrix-end matrix syntax (I'm not posting my attempts at doing it b/c they've all failed completely so I'm sure I'm doing something fundamental, not nuanced, wrong). I think I'm doing something wrong especially in trying to get the syntax to use a data file rather than a matrix written into the syntax (all examples I found used the latter).

I'm sure there's a straightforward way of doing this that someone can help me out with! :-)

Thanks,
Mike

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[David Marso]

GET FILE "your matrix which is saved as an SPSS system file".
MATRIX.
GET X /FILE=* /VARIABLES ALL.
COMPUTE nr=NROW(X).
COMPUTE nc=NCOL(X).
PRINT nr.
PRINT nc.
COMPUTE SUMVAR=0.
COMPUTE SUMCOV=0.
LOOP I=1 TO nr.
+  LOOP J=1 TO I.
+    DO IF (I EQ J).
+      COMPUTE SUMVAR=SUMVAR+X(I,J).
+    ELSE.
+      COMPUTE SUMCOV=SUMCOV+X(I,J).
+    END IF.
+  END LOOP.
END LOOP.
PRINT SUMVAR.
PRINT SUMCOV.
END MATRIX.

Michael Parent wrote

Hi, all,
I'm struggling with an spss syntax issue. I need to perform some basic analyses on a saved covariance matrix. I can't for the life of me get this to work under normal syntax or within matrix-end matrix syntax.

I've got covariance matrices (of varying size). They're saved as spss saved data files; the only data in the files are the actual covariance matrix data (variables labelled var1, var2, …vark, in the spss variable view but NOT as a column in the data set). For the purposes of what I need to get, I don't actually need the N.

ALL I need to do is get the sum of variances (i.e., diagonal elements) and sum of covariances (i.e., one set of the symmetrical half-diagonal elements) from the data sets (I have a few data sets but doing them one at a time is fine). I can't get it to work at all trying syntax commands or using matrix-end matrix syntax (I'm not posting my attempts at doing it b/c they've all failed completely so I'm sure I'm doing something fundamental, not nuanced, wrong). I think I'm doing something wrong especially in trying to get the syntax to use a data file rather than a matrix written into the syntax (all examples I found used the latter).

I'm sure there's a straightforward way of doing this that someone can help me out with! :-)

Thanks,
Mike

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[David Marso]

Here is another version which exploits built in functions of MATRIX language.
MATRIX.
GET X /FILE=* /VARIABLES ALL.
COMPUTE SUMVAR=TRACE(X).
COMPUTE SUMCOV=(MSUM(X)-SUMVAR)/2.
PRINT SUMVAR.
PRINT SUMCOV.
END MATRIX.