scoringRules (version 1.0.1)

GDP data: Data and forecasts for US GDP growth

Description

Historical data and forecast distributions for the growth rate of US gross domestic product (GDP). The forecasts are generated from a Bayesian Markov Switching model as described in Section 5 of Krueger et al (2020).

Format

gdp is a data frame which contains the real-time data set used in Section 5 of KLTG (2020), with the following columns:

• dt - date in question (e.g., "2013Q2" for the second quarter of 2013)

• vint - data vintage (i.e., the date at which the realization was recorded); same format as dt

• val - value of the GDP growth rate

gdp_mcmc is a list, whereby each element is a data frame. gdp_mcmc$forecasts contains the simulated forecast distributions. There are 20 columns (corresponding to quarters 2008:Q1 to 2012:Q4) and 5.000 rows (corresponding to simulation draws). gdp_mcmc$actuals contains the actual observations. There are 20 columns (again corresponding to quarterly dates) and a single row.

Details

The realizations in gdp_mcmc$actuals are also contained in gdp, based on the second available vintage for each date. For example, gdp_mcmc$actuals$X2008Q1 is the entry in gdp for which dt == "2008Q1" and vint == "2008Q3".

References

Krueger, F., Lerch, S., Thorarinsdottir, T.L. and T. Gneiting (2020): `Predictive inference based on Markov chain Monte Carlo output', International Statistical Review, forthcoming. tools:::Rd_expr_doi("10.1111/insr.12405")

Examples

if (FALSE) {

# Load data
data(gdp_mcmc)

# Histogram of forecast draws for 2012Q4
fc_draws <- gdp_mcmc$forecasts[, "X2012Q4"]
hist(fc_draws, main = "Forecast draws for 2012:Q4", xlab = "Value")

# Add vertical line at realizing value
rlz <- gdp_mcmc$actuals[, "X2012Q4"]
abline(v = rlz, lwd = 3)

# Compute CRPS for this forecast case
crps_sample(y = rlz, dat = fc_draws)

}