The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.
mbh_filter()’s automatic knot count is now capped at
250 (min(max(20, floor(n / 2)), 250)). Series of 500
observations or fewer are unaffected; the cap only bounds the B-spline
basis for long or high-frequency inputs, where extra knots inflate
memory and runtime without adding flexibility (in a P-spline the
difference penalty, not the knot count, controls smoothness).d = "auto" default is calibrated from the MAD
of the HP cyclical residual (not first differences), and the default
learning rate is nu = 0.1.hp_filter(),
hamilton_filter(), bhp_filter(),
mbh_filter()). The new boot_iter and
block_size arguments add $trend_lower /
$trend_upper to the result: a 95% normal-approximation band
(trend ± 1.96 * sd) built from a Circular Block Bootstrap
of the cycle, with each replicate refit by the same estimator as the
base fit.autoplot() method for macrofilter
objects (ggplot2): draws the observed series, the estimated trend, and
the confidence ribbon when present, with the time axis reconstructed
from the stored temporal identity.mbh_filter() gains hp_lambda to control
the HP-based auto-calibration of the Huber threshold d when
the input is a plain numeric vector whose true frequency is not
annual.hp_filter() and
bhp_filter() with boot_iter > 0 (and the
base bHP fit), with bit-identical results.d = "auto" calibration in mbh_filter()
now uses the MAD of the HP cyclical residual (output-gap scale) instead
of mad(diff(y)), and reports the chosen value via a
message().c("macrofilter", "list") and store the temporal identity
(meta$ts_class, meta$tsp,
meta$idx) so trend, cycle and bands can all be mapped back
to dates for plotting.boot_iter, block_size, the end-point fan and
the Hamilton conditional band.mbh_filter() documents the
mstop–d interaction (reducing
mstop on long log-level series under-smooths the trend);
hamilton_filter() documents the conditional bootstrap band
behaviour.These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.