R/rescale.R

In djmorris1989/onetmappinguk: Map US O-NET to the UK SOC-2010

#' Rescale O-NET skills
#'
#' This function adjusts the skills data. Some skills were valued by job incumbents, others by occupational experts/analysts. As job incumbents are
#' likely to systematically over-value skill level/importance, the raw skill measures in O-NET are adjusted to account for this. Each skill level
#' and importance measure is regressed on occupation fixed effects, year fixed effects, and a dummy variable indicating if the occupation was
#' rate by job incumbents. The value of this coefficient is subtracted from the skill measure reported by job incumbents.
#'
#' @param p3
#' - The object produced by `phase3_onet()`
#'
#' @return A list two dataframes of panels for each of the O-NET domains
#' - `skill`
#' - `skill.raw`
#'
#' @export

rescale_skills <- function(p3=NULL) {
  skill.raw <- data.frame(p3[["skills.cleaned"]]) %>%
    dplyr::mutate(Incumbent = if_else(Incumbent==0,0,1))
  skill     <- data.frame(p3[["skills.cleaned"]]) %>%
    dplyr::mutate(Incumbent = if_else(Incumbent==0,0,1))

  # LOOP OVER ALL SKILLS
  for (i in 5:74) {
    j <- i - 4
    #print(paste0("Phase 4 of 6: Now Running Skill Regression ",j," of ",70))

    skill$depvar <- skill[,i]
    # fixed effects regression on incumbent dummy, year dummies
    model <- plm(depvar ~ Incumbent + as.factor(year),
                 data=skill,
                 na.action = na.omit,
                 index = c("onetsoc2010code","year"),
                 model="within")

    # subtract the incumbent effect from the data value
    skill[,i] <- skill[,i] - skill[,"Incumbent"]*coef(model)["Incumbent"]
  }
  skill <- skill[,c(1:74)]

  return.list <- list(skill=skill,
                      skill.raw=skill.raw)

  return(return.list)
}




#' Rescale O-NET abilities
#'
#' This function adjusts the abilities data. Some abilities were valued by job incumbents, others by occupational experts/analysts. As job incumbents are
#' likely to systematically over-value ability level/importance, the raw skill measures in O-NET are adjusted to account for this. Each abilities level
#' and importance measure is regressed on occupation fixed effects, year fixed effects, and a dummy variable indicating if the occupation was
#' rate by job incumbents. The value of this coefficient is subtracted from the abilities measure reported by job incumbents.
#'
#' @param p3
#' - The object produced by `phase3_onet()`
#'
#' @return A list two dataframes of panels for each of the O-NET domains
#' - `abilities`
#' - `abilities.raw`
#'
#' @export

rescale_abilities <- function(p3=NULL) {

# No incumbent ratings for abilities

  abilities.raw <- data.frame(p3[["abilities.cleaned"]]) %>%
    dplyr::mutate(Incumbent = if_else(Incumbent==0,0,1))
  abilities     <- data.frame(p3[["abilities.cleaned"]]) %>%
    dplyr::mutate(Incumbent = if_else(Incumbent==0,0,1))

  return.list <- list(abilities=abilities,
                      abilities.raw=abilities.raw)

  return(return.list)
}





#' Rescale O-NET activities
#'
#' This function adjusts the activities data. Some activities were valued by job incumbents, others by occupational experts/analysts. Each activities level
#' and importance measure is regressed on occupation fixed effects, year fixed effects, and a dummy variable indicating if the occupation was
#' rate by job incumbents. The value of this coefficient is subtracted from the activities measure reported by job incumbents.
#'
#' @param p3
#' - The object produced by `phase3_onet()`
#'
#' @return A list two dataframes of panels for each of the O-NET domains
#' - `activities`
#' - `activities.raw`
#'
#' @export

rescale_activities <- function(p3=NULL) {
  activities.raw <- data.frame(p3[["activities.cleaned"]]) %>%
    dplyr::mutate(Incumbent = if_else(Incumbent==0,0,1))
  activities     <- data.frame(p3[["activities.cleaned"]]) %>%
    dplyr::mutate(Incumbent = if_else(Incumbent==0,0,1))

  # LOOP OVER ALL ACTIVITIES
  for (i in 5:86) {
    j <- i - 4
    #print(paste0("Phase 4 of 6: Now Running Work Activity Regression ",j," of ",82))

    activities$depvar <- activities[,i]
    # fixed effects regression on incumbent dummy, year dummies
    model <- plm(depvar ~ Incumbent + as.factor(year),
                 data=activities,
                 na.action = na.omit,
                 index = c("onetsoc2010code","year"),
                 model="within")

    # subtract the incumbent effect from the data value
    activities[,i] <- activities[,i] - activities[,"Incumbent"]*coef(model)["Incumbent"]
  }
  activities <- activities[,c(1:86)]

  return.list <- list(activities=activities,
                      activities.raw=activities.raw)

  return(return.list)
}





#' Rescale O-NET work contexts
#'
#' This function adjusts the skills data. Some skills were valued by job incumbents, others by occupational experts/analysts. As job incumbents are
#' likely to systematically over-value skill level/importance, the raw skill measures in O-NET are adjusted to account for this. Each skill level
#' and importance measure is regressed on occupation fixed effects, year fixed effects, and a dummy variable indicating if the occupation was
#' rate by job incumbents. The value of this coefficient is subtracted from the skill measure reported by job incumbents.
#'
#' @param p3
#' - The object produced by `phase3_onet()`
#'
#' @return A list two dataframes of panels for each of the O-NET domains
#' - `context`
#' - `context.raw`
#'
#' @export

rescale_contexts <- function(p3=NULL) {
  context.raw <- data.frame(p3[["contexts.cleaned"]]) %>%
    dplyr::mutate(Incumbent = if_else(Incumbent==0,0,1))
  context     <- data.frame(p3[["contexts.cleaned"]]) %>%
    dplyr::mutate(Incumbent = if_else(Incumbent==0,0,1))

  # LOOP OVER ALL CONTEXTS
  for (i in 5:59) {

    context$depvar <- context[,i]
    # fixed effects regression on incumbent dummy, year dummies
    model <- plm(depvar ~ Incumbent + as.factor(year),
                 data=context,
                 na.action = na.omit,
                 index = c("onetsoc2010code","year"),
                 model="within")

    # subtract the incumbent effect from the data value
    context[,i] <- context[,i] - context[,"Incumbent"]*coef(model)["Incumbent"]
  }
  context <- context[,c(1:59)]

  return.list <- list(context=context,
                      context.raw=context.raw)

  return(return.list)
}