R/CalculateTrend.R

In OHI-Science/ohicore: Ocean Health Index calculation package

#' CalculateTrend
#' 
#' Used in conf/functions.R to calculate the "trend" dimension of OHI scores.  Trend is 
#' calculated as the proportional change in status across time using a linear regression model.
#' Typically five year periods are used, but this can be changed if other periods 
#' are more appropriate.  For example, for some climate variables, longer time
#' periods might be better to assess long-term climate change.
#' 
#' @param status_data  data frame that includes variables: rgn_id (or, region_id), year, status
#' @param trend_years  list of the years used to calculate trend, usually five years (e.g., 2001:2005)
#'
#' @return Returns an internally used dataframe with region_id, score (i.e., calculated trend values),
#'  and dimensionan (i.e., "trend")
#'#' 
#'
#' @keywords ohicore
#' @examples
#' \dontrun{
#' } 
#' @export

CalculateTrend <- function(status_data, trend_years=trend_years){   
  
  if(sum(grepl("rgn_id", names(status_data))>0)){
    names(status_data)[which(names(status_data)=="rgn_id")] <- "region_id"
  }
  
  if(sum(grepl("scenario_year", names(status_data)) > 0)) {
    names(status_data)[which(names(status_data) == "scenario_year")] <- "year"
  }
  
  status_data <- status_data %>%
    dplyr::select(region_id, year, status) %>%
    dplyr::filter(year %in% trend_years) %>%
    dplyr::filter(!is.na(status)) %>%
    unique()
  
  adj_trend_year <- min(trend_years)
  
  r.trend = status_data %>%
    dplyr::group_by(region_id) %>%
    dplyr::do(mdl = lm(status ~ year, data=.),
       adjust_trend = .$status[.$year == adj_trend_year]) %>%
    dplyr::summarize(region_id, score = ifelse(coef(mdl)['year']==0, 0, coef(mdl)['year']/adjust_trend * 5)) %>%
    dplyr::ungroup() %>%
    dplyr::mutate(score = ifelse(score>1, 1, score)) %>%
    dplyr::mutate(score = ifelse(score<(-1), (-1), score)) %>%
    dplyr::mutate(score = round(score, 4)) %>%
    dplyr::mutate(dimension = "trend") %>%
    dplyr::select(region_id, score, dimension)
  
  return(r.trend)
}