ar.gm(x, order=1, wsave=T, effgm=<<see below>>, effloc=<<see below>>,
        b=T, c=<<see below>>, chr=<<see below>>, cbr=<<see below>>,
        iterh=4, iterb=1)

x:

univariate time series or a real vector. Missing values are not allowed.

order:

integer giving the order of the autoregression.

wsave:

logical parameter determining which array of weights to save. If TRUE, save wsmall; if FALSE, save wbig.

effgm:

parameter giving the desired asymptotic efficiency of gm-estimates of AR coefficients for Gaussian data (based on first-order AR theory). The default is 0.87 or, if c, chr, cbr, and/or effloc are specified, the value is determined by these constants. If effgm is specified, it determines the value of c. This is ignored if iterh = iterb = 0.

effloc:

parameter giving the desired asymptotic efficiency of the m-estimate of location used for centering. This efficiency is also a component in effgm. The default is 0.96 or, if chr and cbr are specified, the value determined by them. If effloc is specified, it determines chr and cbr. This is ignored if iterh = iterb = 0.

b:

logical parameter determining whether a bisquare or Huber psi function is to be used to form wbig weights. If TRUE, a bisquare is used; if FALSE, a Huber is used.

c:

constant used for the psi function for the wbig weights, The default is 4 or the value determined by effgm and effloc if specified. If both effgm and c are specified, the value specified for effgm is used to determine c and the input c is ignored.

chr:

constant used for the Huber psi for the wsmall weights. The default is 1.5 or the value determined by effloc if effloc is specified. Input chr is ignored if effloc is specified. The same chr is used for the estimate of location used to center the data.

cbr:

constant used for the bisquare psi for the wsmall weights. The default is 5 or the value determined by effloc if effloc is specified. Input cbr is ignored if effloc is specified. The same cbr is used for the estimate of location.

iterh:

number of iterations with Huber psi for location and gm estimate residuals. Use iterh = 0 to do least squares.

iterb:

number of iterations with bisquare psi for location and gm estimate residuals. Use iterb = 0 to do least squares.

ar:

vector of length order containing the gm estimates of the AR coefficients.

sinnov:

vector of the innovations scale estimates for the AR models of orders 1 through order.

chat:

an order by order estimated covariance matrix of the process.

rmu:

robust location estimate for x, the sample mean if iterh = iterb = 0.

sd:

robust scale estimate for x, the standard deviation if iterh = iterb = 0.

effgm:

the input, computed or default value of effgm.

effloc:

the input, computed or default value of effloc.

c:

the input, computed or default value of c.

chr:

the input, computed or default value of chr.

bh:

the constant, determined by chr, used in computing Huber Proposal 2 scale estimates.

cbr:

the input, computed or default value of cbr.

bb:

the constant, determined by cbr, used in computing Huber Proposal 2 scale estimates.

w:

time series or vector of weights. These are wsmall if wsave is TRUE, and wbig if wsave is FALSE.

Martin, R. D. (1981). Robust methods for time series. In Applied Time Series Analysis II. D. F. Findley, ed. Academic Press, New York. pp. 683-759. Time Series chapter.

robar <- ar.gm(bicoal.tons,2)