RCT-A subset machine/human relative performance by question group

chart-model.R

library("tidyverse")

arima_qs <- readr::read_csv("/path/to/forecast_rct_a_subset.csv", 
  col_types = cols(user_id = col_integer())) %>%
  mutate(Forecaster = factor(Forecaster))

# Add # of questions to each Data source label
arima_qs <- arima_qs %>%
  group_by(Data_source) %>%
  mutate(by_Data_source_n_ifps = length(unique(ifp_id))) %>%
  ungroup() %>%
  mutate(Data_source = paste0(Data_source, " (", by_Data_source_n_ifps, ")"))
  
mdl3 <- lm(brier ~ -1 + Data_source + Data_source:Forecaster, data = arima_qs)
summary(mdl3)