CORR <data>,L computes the means, standard deviations and correlations of all active variables and observations in <data>. Specifications like VARS, MASK, CASES and IND may be used for selecting variables and observations. Observations with missing values at least in one of the active variables are omitted. For incomplete data CORRMV can be used. The results will appear from line L onwards. If L is missing, results are not displayed in the edit field. If an output file (see OUTPUT) is selected, the results are appended to the end of that file. Printing of results may be completely left out by entering RESULTS=0. In any case the results (in full precision) are saved in matrix files CORR.M correlations MSN.M col.1=means col.2=stddevs col.3=number of observations on the current data disk. Their names can be changed by the OS command >REN, for example. Many operations in multivariate analysis and linear models can use matrix files CORR.M and MSN.M as a basis for their work. /COV computes the covariance matrix COV.M from current CORR.M and MSN.M. All the computations are carried out in double precision. When computing the moments the values of the first observation are used as temporary means. If a variable is a constant, no error message is given, but all the correlations (including the diagonal element, which normally is 1) are set to 0 for that variable. If specification RESULTS=-1 is given, only means and standard deviations (and matrix MSN.M) are computed. Specification RESULTS=0 cancels saving of results in the output file but the matrix files CORR.M and MSN.M are computed even in this case. When the number of variables is high, the best way to study the corre- lation matrix obtained by CORR is to activate /MATSHOW CORR.M or /MATSHOW CORR.M,12.123 (for example). The correlation matrix can also be loaded into the edit field in a partitioned form with significant elements highlighted by /LOADCORR Before activating /LOADCORR it is usually necessary to increase the # of lines and shadow lines by REDIM. LOADM is the general operation used by /MATSHOW and /LOADCORR. L = LINREG M = Multivariate analysis S = General information on statistical operations