predict_alarms: Predict alarms
In PierreMasselot/hhws: Functions for heat-health warning systems

Description Usage Arguments Details Value References Examples

View source: R/find.threshold.R

Given predefined weights and thresholds, predict which of the indicators values are alarms.

1	predict_alarms(indicators, alpha, s, y = NA, r = 1)

`indicators`	List of matrices containing values of the lagged covariates to test whether they are alarms. Can also be a matrix in the case of a single indicator.
`alpha`	List of vectors giving the weightings used to construct indicators. Must have the same length as `indicators` and each vector must have the same length as `ncol` of each matric in `indicators`.
`s`	Vector of thresholds. Must have the same length as `indicators`.
`y`	Optional vector of corresponding response value. Must have the same length as the number of rows of each matrix in `indicators`.
`r`	Positive integer. Number of consecutive values below threshold following an alarm to end the episode.

Extracts the indices in indicators that correspond to an alarm according to the given weigths and thresholds. If y is given, the corresponding values of the response e.g. over-mortality are also returned. In addition, episodes are formed by consecutive alarms (or by alarms occurring with time differences) lower or equal than r.

A data.frame object containing the indices, episode number, and the values of both the response and indicators for all alarms found.

Chebana F., Martel B., Gosselin P., Giroux J.X., Ouarda T.B.M.J., 2013. A general and flexible methodology to define thresholds for heat health watch and warning systems, applied to the province of Quebec (Canada). International journal of biometeorology 57, 631-644.

  library(dlnm)
  data(chicagoNMMAPS)
  x <- chicagoNMMAPS$death
  dates <- as.POSIXlt(chicagoNMMAPS$date)
  n <- nrow(chicagoNMMAPS)

  # Compute over-mortality
  om <- excess(x, dates = dates, order = 15)
  
  # Extract all days for which om is above 40%
  epis <- episodes(om, u = 40)

  # Prepare indicator based on temperature until lag 2
  indic <- matrix(NA, nrow = n, ncol = 3)
  indic[,1] <- chicagoNMMAPS$temp  # lag 0
  indic[,2] <- c(NA, chicagoNMMAPS$temp[-n]) # Lag 1
  indic[,3] <- c(NA, NA, chicagoNMMAPS$temp[1:(n-2)]) # lag 2

  # Evaluate different threshold/indicators based on these episodes
  tested <- find.threshold(indic, epis, u.grid = 20:35, thinning = "episodes",
     order.result = "Episodes_found")

  # Choose a result and predict
  final <- tested[1,]
  predict_alarms(indic, final[1:3], s = final[4], y = om)