
Build a calendar from a sample of another's timeslices
Source:R/solve_by_sample.R
sample_calendar.RdReturns a calendar holding only the chosen timeslices of cal, with the
weights that make it annualise: a sample covering half the year weighs each
of its slices twice. Used to solve a model, or one of its milestone years,
on part of the calendar rather than all of it.
Arguments
- cal
a
calendarto sample.- slices
the timeslices to keep. Any level: a leaf set, or coarser members with every finer slice beneath them.
- counts
optional named integer vector of repeat counts per timeslice, for a draw with replacement.
- name
name for the sampled calendar. A calendar assigned to particular years with
setCalendar()needs one, since the horizon refers to it by name.
Details
A sample keeps the parent's timeslice NAMES, its timeframe levels and its
share per slice. Only membership and weight change, so a model keyed on
the parent's labels still matches, and sum(share * weight) is 1 per level
as in any calendar.
At least two members must survive at every level below the year. Dropping a
level instead re-derives the leaf names – WIN_h00 becomes h00 – after
which the model's data matches nothing and the result solves to zero without
complaint. That is refused rather than attempted, naming the level at fault.
counts carries a bootstrap repeat: a slice drawn twice cannot be a
repeated row, since a calendar holds each timeslice once, so the repetition
becomes a scaled share.
timescales::filter_calendar() is the same operation on a
timescales::Calendar, selecting by timeframe and label and recording the
parent and the coverage in the object's metadata. This one works on
energyRt's own calendar class, which is why it exists separately.
See also
calendar_samples() to plan a set of draws,
solve_by_sample() to solve them as replicates, and setCalendar() to
give one to particular milestone years.
Examples
cal <- calendars$s4_h24
# two of the four seasons, with every hour beneath them
smp <- sample_calendar(cal, grep("^(WIN|SUM)", cal@timeslice_share$timeslice,
value = TRUE), name = "half")
smp@year_fraction
#> [1] 0.005479452
# share is the parent's; weight annualises
head(smp@timeslice_share, 3)
#> timeslice share weight
#> <char> <num> <num>
#> 1: ANNUAL 1.000000000 1.0
#> 2: WIN 0.002739726 182.5
#> 3: SUM 0.002739726 182.5