decomplm

A package for various type of linear model decomposition.

Installation

You can install decomplm from github using the devtools package

install.packages("devtools")
require(devtools)
install_github('yoruhyde/decomplm')
library(decomplm) # load package

Example

Usually, work directory, data_var and cs_var change everytime you setting up a new project. Make sure double check them before running the code.

setwd("C:\\Users\\yuemeng1\\Desktop\\code\\decomplm\\data") # set work directory
data=fread("input_data.csv") # load data
var=fread("input_var.csv") # load variable coefficients
layer=fread("input_layer.csv") # load model structure

date_var="week" # date variable name in input_data
cs_var=c("cs") # variable names for cross section in input_data; support multiple dimensions

Then you can run the code below to get the final decomposition of the main model. The code would choose the way that you specified in the file input_layer.csv - column if_direct

if.direct=unique(layer$is_direct[!is.na(layer$is_direct)])

if (length(if.direct)==1) {
  if(if.direct==1) {
    result=direct_weight(is.output.final=T,model_name=NULL)
  } else {
    result=seperate_decomp(is.output.final=T,model_name=NULL)
  }
} else {
  result=mixed_approach(is.output.final=T,model_name=NULL)
}



runApp(result$app) # activate output panel


drpmcain/decomp documentation built on May 15, 2019, 2:51 p.m.