tsoutliers (version 0.3)

calendar.effects: Calendar Effects

Description

This function creates regressor variables for trading day, Easter and leap year effects over the sample period where the input time series is sampled.

Usage

calendar.effects(x, trading.day = TRUE, easter = 6, 
  leap.year = FALSE, holidays = NULL, easter.date = FALSE)

Arguments

x

a monthly time series.

trading.day

logical. If TRUE, trading day regressor variable is returned.

easter

numeric. The number of days before Easter over which the Easter effect spans. If it is set to zero Easter variable is not returned.

leap.year

logical. If TRUE, leap year regressor variable is returned.

holidays

an optional numeric vector of the same length as x containing the number of holidays for each time period.

easter.date

logical indicating whether the date of Easter should be returned.

Value

• A mts matrix containing the selected calendar effects by columns.

  If easter.date is TRUE a list is returned containing the mts matrix of calendar effects as well as the dates of Easter for each year of the sample.

Details

Let $wd$ be the number of working days in a given month and $nwd$ the number of non-working days. The trading day variable at time $t$ is built as follows:

$$wd = wd - holidays$$ $$nwd = nwd + holidays$$ $$td_t = wd - (5/2) \times nwd$$

By default, working days are the days from Monday to Friday and non-working days are Saturdays and Sundays. If there are additional non-working days they can be defined in argument holidays. For example, if the 1st of February is a local holiday, the user can define a variable containing zeros for all periods except for the periods related to February where the 1st of February falls within a working day (Monday to Friday); these data are set to one so that they are considered as non working days.

Easter effect is defined as the proportion of days before Easter (by default easter = 6) that lie in March and April, respectively for each month. It contains zeros for the remaining months.

The leap year is a vector of zeros for all months except February, where the variable takes on the value $0.75$ if the year is a leap year and $-0.25$ otherwise.

Examples

# display calendar effects for a sample span period
# no data are actually necessary in the input series 
# since calendar effects are concerned only with the dates
# at which the data are sampled
x <- ts(frequency = 12, start = c(1980, 1), end = c(2000, 12))
ce <- calendar.effects(x, leap.year = TRUE)
colnames(ce)
plot(ce, main = "calendar effects")
# Easter days for each year
calendar.effects(x, easter.date = TRUE)$easter