forecasthistogram

About

This R package contains functions to handle and study subjective probabilities (‘forecast histograms’) as collected in several economic surveys. The package accompanies the paper ‘Quantifying subjective uncertainty in survey expectations’ by Fabian Krüger and Lora Pavlova, open access available at https://doi.org/10.1016/j.ijforecast.2023.06.001 (International Journal of Forecasting).

The package is presently available via GitHub only.

# Use next two lines to install package from GitHub
# library(remotes)
# install_github("FK83/forecasthistogram")

# Load forecasthistogram package
library(forecasthistogram)
# load magrittr package for pipe operator
library(magrittr)

Example

The following code generates a forecast histogram object, plots it, and computes several summary measures.

# Vector of upper bin limits used in the NY Fed's Survey of Consumer Expectations
ub <- c(-12, -8, -4, -2, 0, 2, 4, 8, 12, Inf)
# Example vector of probabilities
p <- c(rep(0, 5), c(.15, .2, .5, .1, .05))
# Make forecast histogram object and quantify it (see ?quantify)
f <- forecasthistogram(p, ub) %>% quantify
# Plot histogram
plot(f)

# Compute mean and quantiles
mean(f)

## [1] 5.249223

quantile(f, probs = c(.1, .25, .5, .75, .9))

##      10%      25%      50%      75%      90% 
## 1.795743 3.048362 4.898946 7.106687 9.219803

# Compute ERPS measure proposed by Krüger and Pavlova (2022)
erps(f)

## [1] 0.53

# Alternatively: Computation based on original probabilities
sum(cumsum(p)*(1-cumsum(p)))

## [1] 0.53