Incompatibilities with S-PLUS 3.3.
Some are intentional, others are bugs.

spectrum
--------

Autoregressive spectrum estimation not implemented (bug).

spec.pgram
----------

The periodogram at zero is not returned for the raw periodogram and is
not used when calculating the smoothed periodogram (feature)

The Fourier frequencies are different for series of odd length (feature)
This has a knock-on effect on the bandwidth.

Formula for degrees of freedom is correct only for case of no padding
and no tapering (bug). This has a knock-on effect for the confidence
intervals in plot.spec.

Padding. The series will be automatically padded to a length which is a
highly composite number in order to aid the fft algorithm. The padding
behaviour is different from S-PLUS (feature: I can't work out the
S-PLUS padding behaviour).

plot.spec
---------

Implemented as a method for the generic function plot (feature).

Confidence interval is calculated differently (feature).

plot.acf
--------

Implemented as a plot method for objects of class "acf" (feature).

ar
--

Burg's algorithm not implemented (bug)

ar.yw
-----

ar coefficients disagree with S-PLUS for multivariate time
series. (feature, I think this is a bug in S-PLUS.  The algorithm
for involves alternating between the forward and backward Yule-Walker
equations, which have different solutions in the multivariate case.
If there are m series, S-PLUS gives the solution to the (forward)
Yule-Walker equations up to order m, but the solution to the backward
equations for higher orders. This doesn't seem to affect the prediction
variance or the partial correlation.)

aic is different (feature). S-PLUS uses the conditional likelihood,
conditioning on the first p observations for an ar(p) process, so the
likelihood ratio between models of different order - and hence the AIC  -
is not meaningful.  This is a known "feature" of arima.mle in S-PLUS. It
looks like ar does the same thing. I have used formula 17.29 from the
"S-PLUS 3.3 guide to statistical and mathematical analysis" for the
AIC. 
