FCORRELATION procedure

Forms the correlation matrix for a list of variates (R.W. Payne).

Options

PRINT = strings

Printed output (correlations, test); default corr

METHOD = string

Type of test to make (against zero) for the correlations (twosided, greater, lessthan); default twos

WEIGHTS = variate

Provides weights for the units of the variates; default * assumes that they all have weight one

CORRELATIONS = symmetric matrix

Saves the correlations

PROBABILITIES = symmetric matrix

Saves the test probabilities

NOBSERVATIONS = scalars

Saves the number of observations from which the correlations have been calculated

Parameter

DATA = variates

Variates for which the matrix is to be calculated

Method

FCORRELATION forms the correlation matrix for a set of variates which should be listed by the DATA parameter. The WEIGHTS option can provide a variate of weights for the units of the variates; by default these are all assumed to have weight one.

Printed output is controlled by the PRINT option with settings:

correlations

prints the correlation matrix;

tests

prints tests for the correlations.

By default PRINT=correlation.

The METHOD option indicates the type of test to be done, with settings:

twosided

for a two-sided test of the null hupothesis that that the correlation is zero;

greaterthan

for a one-sided test of the null hypothesis that the correlation is not greater than zero;

lessthan

for a one-sided test of the null hypothesis that the correlation is not less than zero.

Tests cannot be produced if there are fewer than two observations.

The correlation matrix can be saved using the CORRELATIONS option, the (symmetric) matrix of test probabilities can be saved using the PROBABILITIES option, and the number of observations upon which it is based can be saved using NOBSERVATIONS option.

Options: PRINT, METHOD, WEIGHTS, CORRELATIONS, PROBABILITIES, NOBSERVATIONS.

Parameter: DATA.

Method

A SSPM structure is set up for the DATA variates, and its values are formed using the FSSPM directive. The corrected sums of squares and products are divided by the residual degrees of freedom to give the variance covariance matrix, and the CORRMAT finction is used to convert this to a correlation matrix. The test probabilities are obtained using the PRCORRELATION procedure.

Action with RESTRICT

FCORRELATIONS takes account of restrictions on any of the variates.