multi_EVPI {decisionSupport}

Expected value of perfect information (EVPI) for multiple variables. This is a wrapper for the empirical_EVPI function. See the documentation of the empirical_EVPI function for more details.

Description

Expected value of perfect information (EVPI) for multiple variables. This is a wrapper for the empirical_EVPI function. See the documentation of the empirical_EVPI function for more details.

Usage

multi_EVPI(mc, first_out_var, write_table = FALSE, outfolder = NA)

## S3 method for class 'EVPI_outputs'
summary(object, ...)

## S3 method for class 'EVPI_outputs'
plot(
  x,
  out_var,
  fileformat = NA,
  outfolder = NA,
  scale_results = TRUE,
  legend_table = NULL,
  output_legend_table = NULL,
  ...
)

Arguments

mc

output table from a Monte Carlo simulation, e.g. as realized with the decisionSupport package

first_out_var

name of the column in the mc table that contains the first output variable. Information Values are computed for variables in all earlier columns.

write_table

boolean parameter indicating whether an output table should be written.

outfolder

folder where the outputs should be saved (this is optional).

object

EVPI_res object (produced with multi_EVPI) as input to the summary function plot.

...

Arguments to be passed to methods, such as graphical parameters (see par).

x

object of class EVPI_outputs as produced with the multi_EVPI function

out_var

name of the output variable to be plotted

fileformat

The file format to be used for the outputs. Currently only NA (for R plot output) and "png" (for a PNG file) are implemented. Note that when this is !NA, the outfolder parameter must point to a valid folder.

scale_results

boolean variable indicating if resulting high numbers should be scaled to avoid numbers in the plot that cannot be read easily. If this is TRUE, numbers are divided by an appropriate divisor and a suffix is added to the number in the plot (e.g. "in millions").

legend_table

a data.frame with two columns variable and label. The variable column should contain the name of the independent variables as listed in the Monte Carlo table. The label column should contain the label to be used for this variable in the EVPI plot.

output_legend_table

a data.frame with two columns variable and label. The variable column should contain the name of the dependent variables as listed in the Monte Carlo table. The label column should contain the label to be used for this variable in the EVPI plot. Note that labels for both dependent and independent variables can be provided in the same table. Then both parameters legend_table and output_legend_table can point to the same table.

Value

invisible list of as many elements as there are output variables in the Monte Carlo table: each element refers to one of the output variables and contains a data.frame with five columns: (1) variable - the input variable names (2) expected_gain - expected gain when project is implemented, without knowing the value of the test variable, equals NA in case there is no variation in the tested variable (3) EVPI_do - the Expected Value of Perfect Information on the respective input variable, if the analysis suggests that the expected value of the decision is likely positive (e.g. the project should be done) (4) EVPI_dont - the Expected Value of Perfect Information on the respective input variable, if the analysis suggests that the expected value of the decision is likely negative (e.g. the project should not be done) (5) the decision whether to implement with the project based on imperfect information

Author(s)

Eike Luedeling, Katja Schiffers

Examples

### In the following example, the sign of the calculation
### is entirely determined by the variable indep1, so
### this should be expected to have a high EVPI. Variable
### indep2 doesn't affect the sign of the output, so it
### should not have information value.

montecarlo <- data.frame(indep1 = rnorm(1000), indep2 = rnorm(1000, 3))
montecarlo[, 'output1'] <- montecarlo[, 'indep1'] * montecarlo[, 'indep2']
montecarlo[, 'output2'] <- (montecarlo[, 'indep1'] * (montecarlo[, 'indep2']) + 10)

results_all <- multi_EVPI(montecarlo,"output1")
summary(results_all)
plot(results_all, "output1")
plot(results_all, "output2")


### In the following example, the sign of the calculation is entirely
### determined by the variable indep1, so this should be expected to have
### a high EVPI. Variable indep2 doesn't affect the sign of the output,
### so it should not have information value.

montecarlo <- data.frame(indep1 = rnorm(1000), indep2 = rnorm(1000, mean = 3))
montecarlo[, 'output1'] <- montecarlo[, 'indep1'] * montecarlo[, 'indep2']
montecarlo[, 'output2'] <- (montecarlo[, 'indep1'] * (montecarlo[, 'indep2']) + 10)

results_all <- multi_EVPI(montecarlo,"output1")
summary(results_all)
plot(results_all, "output1")
plot(results_all, "output2")