GEKS_decomp: Decomposition of GEKS-tornqvist

In MjStansfi/GEKSdecomp:

Description

Take smiliar inputs as the GEKS function and return the decomposition on the multilateral scale.

Usage

GEKS_decomp(times, price, id, weight, custom_time = c(), window_length = NULL)

Arguments

times	vector of the times at which price observations were made. NOTE: times must be of class Date or numeric.
price	vector of log of prices at the given time
id	vector of distinct identification number of consumer goods
weight	vector of expenditure weights used in the regressions
custom_time	either empty (will assume latest periods) or a vector of length two relating to times compared where [1] is the 'start_time' time and [2] is the 'to' time e.g Comparing contribution start_time "1973-04-01" to "1973-05-01" would be c("1973-04-01", "1973-05-01")
window_length	optional. Single number for length of window for the data that regressions are fit on. Note if window_length is present it assumes custom time is latest two periods in window.

Value

A dataframe containing the numeric contribution (contrib) of every price observation in the window, the product of which will (roughly) equal the index movement between the two time periods. p_contrib is the standardised percentage contribution of the observation, the sum of which will equal 1.

Examples

library(TPDdecomp)

contributions <- with(
    turvey,
    GEKS_decomp(times = month,
               price = price,
               id = commodity,
               weight = price*quantity,
               custom_time = as.Date(c("1973-04-30","1973-05-31")),
               window_length = NULL)
 )


str(contributions)
Classes ‘data.table’ and 'data.frame':	5 obs. of  5 variables:
$ from     : Date, format: "1973-04-30" "1973-04-30" "1973-04-30" ...
$ to       : Date, format: "1973-05-31" "1973-05-31" "1973-05-31" ...
$ id       : Factor w/ 5 levels "Apples","Grapes",..: 1 2 3 4 5
$ contrib  : num  1.035 0.998 1.017 1.034 1
$ p_contrib: num  0.411 -0.02 0.206 0.403 0

MjStansfi/GEKSdecomp documentation built on Sept. 30, 2020, 4:21 p.m.