Make a Composed Time Series Function — make_ts

Build a function that returns a product of four configurable elements:

$\bar x$: a mean value
$F_S(t)$: a seasonal pattern
$F_T(t)$: a trend
$F_K(t)$: a shock

The return value is a function of the form:

$F_t(t)$ if form == "t"
$F_t(t, V(t))$ if form == "tV"

The user can pass any vector times and the interval is set to

$t_0 = \mbox{min(times)}$, and
$t_1 = \mbox{max(times)}$.

A normalizing constant is set such that the daily average over the interval $(t_0, t_1)$ is $\bar x$: $$\int_{t_0}^{t_1} F_t(t) dt = \bar x \left(t_1-t_0\right).$$

The component functions are specified by passing F_obj objects (see F_obj):

season_par creates $F_S(t)$ or F_season (eg, using makepar_F_sin)
trend_par creates $F_T(t)$ or F_trend (eg, using makepar_F_spline)
shock_par creates $F_K(t)$ or F_shock (eg, using makepar_F_sharkbite)

Usage

make_ts_function(
  avg = 1,
  season_par = list(),
  trend_par = list(),
  shock_par = list(),
  times = c(0, 365),
  norm_with_shocks = FALSE,
  N = 1,
  form = "tV",
  options = list()
)

Arguments

avg: the average
season_par: seasonality function for
trend_par: trend function parameters
shock_par: trend function parameters
times: normalize $t_0 = \mbox{min(times)}$ to $t_1 = \mbox{max(times)}$
norm_with_shocks: if FALSE, set $F_K(t)=1$ for normalization
N: the length of the return value
form: functional form: "t" returns $F(t)$; "tV" returns $F(t,V)$
options: configurable options

Value

a function