Skip to contents

Temporal disaggregation of a time series by regression models.

Source: R/tempdisagg.R
temporaldisaggregation.Rd

Perform temporal disaggregation of low frequency to high frequency time series by regression models. Models included are Chow-Lin, Fernandez, Litterman and some variants of those algorithms.

Usage

temporaldisaggregation(
  series,
  constant = TRUE,
  trend = FALSE,
  indicators = NULL,
  model = c("Ar1", "Rw", "RwAr1"),
  freq = 4,
  conversion = c("Sum", "Average", "Last", "First", "UserDefined"),
  conversion.obsposition = 1,
  rho = 0,
  rho.fixed = FALSE,
  rho.truncated = 0,
  zeroinitialization = FALSE,
  diffuse.algorithm = c("SqrtDiffuse", "Diffuse", "Augmented"),
  diffuse.regressors = FALSE
)

Arguments

series

The time series that will be disaggregated. It must be a ts object.

constant

Constant term (T/F). Only used with Ar1 model when zeroinitialization=F

trend

Linear trend (T/F)

indicators

High-frequency indicator(s) used in the temporal disaggregation. It must be a (list of) ts object(s).

model

Model of the error term (at the disaggregated level). "Ar1" = Chow-Lin, "Rw" = Fernandez, "RwAr1" = Litterman

freq

Annual frequency of the disaggregated variable. Used if no indicator is provided

conversion

Conversion mode (Usually "Sum" or "Average")

conversion.obsposition

Only used with "UserDefined" mode. Position of the observed indicator in the aggregated periods (for instance 7th month of the year)

rho

Only used with Ar1/RwAr1 models. (Initial) value of the parameter

rho.fixed

Fixed rho (T/F, F by default)

rho.truncated

Range for Rho evaluation (in [rho.truncated, 1[)

zeroinitialization

The initial values of an auto-regressive model are fixed to 0 (T/F, F by default)

diffuse.algorithm

Algorithm used for diffuse initialization. "SqrtDiffuse" by default

diffuse.regressors

Indicates if the coefficients of the regression model are diffuse (T) or fixed unknown (F, default)

Value

An object of class "JD3TempDisagg"

Examples

# retail data, chow-lin with monthly indicator
Y<-rjd3toolkit::aggregate(rjd3toolkit::retail$RetailSalesTotal, 1)
x<-rjd3toolkit::retail$FoodAndBeverageStores
td<-rjd3bench::temporaldisaggregation(Y, indicators=x)
y<-td$estimation$disagg

# qna data, fernandez with/without quarterly indicator
data("qna_data")
Y<-ts(qna_data$B1G_Y_data[,"B1G_FF"], frequency=1, start=c(2009,1))
x<-ts(qna_data$TURN_Q_data[,"TURN_INDEX_FF"], frequency=4, start=c(2009,1))
td1<-rjd3bench::temporaldisaggregation(Y, indicators=x, model = "Rw")
td2<-rjd3bench::temporaldisaggregation(Y, model = "Rw")
mod1<- td1$regression$model