Run scul procedure on placebo pool to create null distribution

Source: R/CreatePlaceboDistribution.R

Runs the scul procedure on the whole placebo pool (using itself as the donor pool; this could be altered to have a custom donor pool) to create null distribution that can be used for inference in later steps.

CreatePlaceboDistribution(
  x.PlaceboPool = SCUL.input$x.PlaceboPool,
  TreatmentBeginsAt = SCUL.input$TreatmentBeginsAt,
  PostPeriodLength = nrow(SCUL.input$y) - SCUL.input$TreatmentBeginsAt + 1,
  PrePeriodLength = SCUL.input$TreatmentBeginsAt - 1,
  NumberInitialTimePeriods = SCUL.input$NumberInitialTimePeriods,
  OutputFilePath = SCUL.input$OutputFilePath,
  CohensDThreshold = SCUL.input$CohensDThreshold,
  TrainingPostPeriodLength = SCUL.input$TrainingPostPeriodLength,
  DonorPoolRestrictionForEachPlacebo = "",
  cvOption = "lambda.1se"
)

Arguments

x.PlaceboPool

A (T by J) data frame containing all products that you wish to include in the placebo distribution Must be sorted by time. Default is SCUL.input$x.PlaceboPool.
TreatmentBeginsAt

An integer indicating which row begins treatment. Default is SCUL.output$TreatmentBeginsAt.
PostPeriodLength

An integer that indicates the length of the post-treatment period. Defailt is calculated using SCUL.input data.
PrePeriodLength

An integer that indicates the length of the pre-treatment period. Defailt is calculated using SCUL.input data.
NumberInitialTimePeriods

An integer that indicates the number of time periods desired in the training data for the first cross-validation run. Default is the stated amount in SCUL.input data.
OutputFilePath

Output file path. Default is SCUL.input$OutputFilePath.
CohensDThreshold

A real number greater than 0, indicating the Cohen's D threshold at which fit is determined to be "poor". The difference is in standard deviation units. Default is SCUL.input$CohensDThreshold.
TrainingPostPeriodLength

The number of timer periods post-treatment for training data. Defaults to all time since treatment begins.
DonorPoolRestrictionForEachPlacebo

This is a customizeable restriction on the placebo pool. In Hollingsworth and Wing (2020) we use it to ensure that no analysis on a placebo series uses donors from the same state. Where state is identified by a pattern in each variable name. Default is no restriction. The placebo pool is indexed by `h`: For a restriction based on the first two characters: "select(-starts_with(substring(names(x.PlaceboPool)[h],1, 2)))" For a restriction based on the last two characters: "select(-ends_with(substring(names(x.PlaceboPool)[h],nchar(names(x.PlaceboPool)[h]) - 2 + 1, nchar(names(x.PlaceboPool)[h]))))"
cvOption

Do you want to use the median CV lambda (cvOption = lambda.median), one that produces the minimum MSE (cvOption = lambda.min), or the largest lambda that produces a MSE within one standard error of the minimum MSE (cvOption = lambda.1se). Default is lambda.1se

Value

list A list of standardized placbo effect sizes