CRPS and fan chart/simulated futures

Machine Learning and Modeling
,
1

Hi,

Question regarding CRPS:
The CRPS averages the quantile scores over all values of p i.e. the whole the forecast distribution. But exactly what is the whole the forecast distribution? Because if I do an autoplot of a fan chart with level = 100, the fan is all the way from the top to the bottom; does this mean the CRPS is a measure for 100% prediction interval's ? Or is the CRPS a measure of the whole distribution given the parameters of the model i.e. a fan chart as one would plot from 10,000 simulated futures?

library(fpp3)

# Re-index based on trading days
google_stock <- gafa_stock |>
  filter(Symbol == "GOOG", year(Date) >= 2015) |>
  mutate(day = row_number()) |>
  update_tsibble(index = day, regular = TRUE)

# Filter the year of interest
google_2015 <- google_stock |> filter(year(Date) == 2015)

# forecast OOS
google_2015 |>
  model(NAIVE(Close)) |>
  forecast(h = 10) |>
  autoplot(google_2015, level = c(80, 95, 100)) +
  labs(title="Google daily closing stock price", y="$US" )

thanks,
Amarjit

2

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.