Next: specarma() Up: Arima Macros Help File Previous: rhatvar()   Contents


rhatcovar(rho,i, j) or rhatcovar(rho,lag:L), REAL vector rho, integers
  i > 0, j > 0, L > 0

Keywords: time domain, autocorrelation
rhatcovar(rho [,lag:L]), where rho is a REAL vector of auto-
correlations, computes n*Var[rhohat], where Var[rhohat] is the large
sample covariance matrix of the sample autocorrelation function
rhohat, computed using Bartletts' formula.

The result is a REAL L by L matrix.  The default value of L is
(maximum lag for rho)/2.

When rho[1] == 1, rho[k] is assumed to contain the lag k-1 auto-
correlation; otherwise, rho[k] is assumed to contain the lag k
autocorrelation.  Thus if gamma is a REAL vector containing the
autocovariance function with gamma[1] = Var[X(t)], both
rhatvar(gamma/gamma[1]) and rhatvar(gamma[-1]/gamma[1]) return the same

rhatcovar(rho,i,j) returns n*covar{rhohat(i),rhohat(j)}, that is,

See also rhatvar().

Gary Oehlert 2003-01-15