Overview (PARTIAL CORR command)

PARTIAL CORR produces partial correlation coefficients that describe the relationship between two variables while adjusting for the effects of one or more additional variables. PARTIAL CORR calculates a matrix of Pearson product-moment correlations. PARTIAL CORR can also read the zero-order correlation matrix as input. Other procedures that produce zero-order correlation matrices that can be read by PARTIAL CORR include CORRELATIONS, REGRESSION, DISCRIMINANT, and FACTOR.

Options

Significance Levels. By default, the significance level for each partial correlation coefficient is based on a two-tailed test. Optionally, you can request a one-tailed test using the SIGNIFICANCE subcommand.

Statistics. In addition to the partial correlation coefficient, degrees of freedom, and significance level, you can use the STATISTICS subcommand to obtain the mean, standard deviation, and number of nonmissing cases for each variable, as well as zero-order correlation coefficients for each pair of variables.

Format. You can specify condensed format, which suppresses the degrees of freedom and significance level for each coefficient, and you can print only nonredundant coefficients in serial string format by using the FORMAT subcommand.

Matrix Input and Output. You can read and write zero-order correlation matrices by using the ­MATRIX subcommand.

Basic Specification

The basic specification is the VARIABLES subcommand, which specifies a list of variables to be correlated, and one or more control variables following keyword BY. PARTIAL CORR calculates the partial correlation of each variable with every other variable that was specified on the correlation variable list.

Subcommand Order

Subcommands can be specified in any order.

Operations

  • PARTIAL CORR produces one matrix of partial correlation coefficients for each of up to five order values. For each coefficient, PARTIAL CORR prints the degrees of freedom and the significance level.
  • This procedure uses the multithreaded options specified by SET THREADS and SET MCACHE.

Limitations

  • A maximum of 25 variable lists on a single PARTIAL CORR command is allowed. Each variable list contains a correlation list, a control list, and order values.
  • A maximum of 400 variables total can be named or implied per PARTIAL CORR command.
  • A maximum of 100 control variables is allowed.
  • A maximum of 5 different order values per single list is allowed. The largest order value that can be specified is 100.