MC.analysis_generic: Analysis of the Monte Carlo simulation (general function)

In stUPscales: Spatio-Temporal Uncertainty Propagation Across Multiple Scales

Analysis of the Monte Carlo simulation (general function)

Description

General function for running the analysis of the Monte Carlo simulation.

Usage

  MC.analysis_generic(x, delta, qUpper, data.det, sim.det, event.ini, event.end, 
              ntick, summ.data = NULL)

Arguments

x	A list of 1, which contains the output of the Monte Carlo simulation as a data.frame with n rows as time steps and the first column is time in format POSIXct and m columns named 1, 2, 3... m, where m is the number of Monte Carlo runs results.
delta	A numeric value that specifies the level of aggregation required in minutes.
qUpper	A character string that defines the upper percentile to plot the confidence band of results, several options are possible "q999" the 99.9th percentile, "q995" the 99.5th percentile, "q99" the 99th percentile, "q95" the 95th percentile, "q50" the 50th percentile. The lower boundary of the confidence band (showed in gray in the output plots) is the 5th percentile in all cases.
data.det	A data.frame that contains the time series of the main driving force of the system to be simulated deterministically, e.g. precipitation. This data.frame should have only two columns: the first one, Time [y-m-d h:m:s] in POSIXct format; the second one, a numeric value equal to the magnitude of the variable.
sim.det	A list of 1 that contains the results of the deterministic simulation, here the output given data.det. The format is the same as data.det

.

event.ini	A time-date string in POSIXct format that defines the initial time for event analysis.
event.end	A time-date string in POSIXct format that defines the final time for event analysis.
ntick	A numeric value to specify the number of ticks in the x-axis for the event time-window plots.
summ.data	A list by default NULL. If provided, the list should contain an output of the MC.analysis function, and the analysis is done again without the calculation of some of the internal variables, therefore the analysis is faster.

Value

A list of length 2:

summ

A list that contains the summary statistics of the Monte Carlo simulation per output variable. Each output variable is summarised by calculating the mean "Mean", standard deviation "sd", variance "Variance", 5th, 25th, 50th, 75th, 95th, 99.5th, 99.9th percentiles "q05", "q25", "q50", "q75", "q95", "q995", "q999", the max "Max", the sum "Sum", time "time", and the deterministic data "p1", all variables as time series.

variance

A data.frame that contains the summary statistics of the variance of the Monte Carlo simulation per output variable.

Author(s)

J.A. Torres-Matallana

See Also

See also setup-class, MC.setup-methods, MC.sim-methods.

Examples

## Creating meta-model
Model <- function(A, B, variable.1, variable.2){
  lum <- A*variable.1 + B*variable.2
}

## Model input and parameter set-up

time <- data.frame(time = seq.POSIXt(from = as.POSIXct("2019-01-01"), 
                                     to = as.POSIXct("2019-01-02"), by = 60*60*6))
data <- cbind(time, data = 25) 
data

new.setup <- setup(id = "MC_1",
                   nsim = 10,
                   seed = 123,
                   mcCores = 1,
                   ts.input = data, 
                   rng = rng <- list(
                     A = 1.25,
                     B = 0.75,
                     variable.1 = c(pdf = "uni", min = 0, max = 4),
                     variable.2 = c(pdf = "uni", min = 2.2, max = 3.2)
                   )
)

str(new.setup)

## Monte Carlo simulation set-up
set.seed(slot(new.setup, "seed"))

new.mc.setup <- MC.setup(new.setup)
str(new.mc.setup)         

## Monte Carlo simulation
output <- data.frame(time = new.mc.setup$ts.input[,1])
output[,2:(new.mc.setup$nsim + 1)] <- NA

for(i in 1:new.mc.setup$nsim){
  for(j in 1:nrow(new.mc.setup$ts.input)){
    
    ## model parameter definition
    A <- new.mc.setup$par$A
    B <- new.mc.setup$par$B
    
    ## model input definition
    variable.1 <- new.mc.setup$par$variable.1[i,j]
    variable.2 <- new.mc.setup$par$variable.2[i,j]
    
    ## model evaluation
    output[j,i+1] <- Model(A, B, variable.1, variable.2)  
  }
}

output <- list(output1 = output)
output  

## Deterministic simulation
# model parameter definition
A <- new.mc.setup$par$A
B <- new.mc.setup$par$B

# model input definition
variable.1.det <- apply(X = new.mc.setup$par$variable.1, MARGIN = 2, FUN = mean)
variable.2.det <- apply(X = new.mc.setup$par$variable.2, MARGIN = 2, FUN = mean)

output.det      <- Model(A, B, variable.1.det, variable.2.det)  
output.det      <- cbind(time, output.det)
output.det      <- list(out1 = output.det)
str(output.det)

## Monte Carlo analysis
delta     <- 60*6 # minutes
qUpper    <- "q95"
event.ini <- data$time[1]
event.end <- data$time[nrow(data)]
ntick     <- 1

analysis <- MC.analysis_generic(x = output, delta = delta, qUpper = qUpper, data.det = data,
                                sim.det = output.det, event.ini = event.ini, event.end = event.end,
                                ntick = ntick)

stUPscales documentation built on Sept. 18, 2023, 9:07 a.m.