covar(x), REAL matrix x with no MISSING values, returns structure(n:sampleSize,mean:xbar,covariance:s) |

covar(x), where x is a REAL matrix with no MISSING values computes the sample mean and variance-covariance matrix of x. It returns structure(n:N, mean:xbar, covariance:s) where N = nrows(x) is the sample size, xbar is a row vector of the sample means of the columns, and s is the sample variance-covariance matrix (with divisor N - 1). Macro covar() is obsolete but is retained for backward compatibility. Essentially the same output can be obtained from tabs(y, count:T, covar:T, mean:T) which returns structure(mean:means, covar:s,count:n). The components of covar() output differ from those of tabs() in three ways: (a) Component names ('covariance' instead of 'covar' and 'n' instead of 'count') (b) The value of 'mean' is a row vector (1 by ncols(x)) for covar() and a (column) vector for tabs() (c) The value of covar() component 'n' is the scalar N, while tabs() component 'count' is rep(N,ncols(x)), with a count for each column of x See also tabs() and groupcovar().

Gary Oehlert 2003-01-15