Skip to contents

Standardize traded event probabilities as an event_prices object

Source: R/event_prices.R
as_event_prices.Rd

as_event_prices() turns a data.frame/tibble holding a time series of traded event probabilities (prediction-market prices, betting quotes, state prices) into a standardized event_prices tibble that all estimation and plotting functions of the package understand.

Usage

as_event_prices(x, ...)

# S3 method for class 'event_prices'
as_event_prices(x, ...)

# S3 method for class 'data.frame'
as_event_prices(
  x,
  time = NULL,
  price = NULL,
  bid = NULL,
  ask = NULL,
  scale = 1,
  discount = 1,
  book = NULL,
  method = c("discount", "overround"),
  clip = ec_default_params()$clip,
  market_id = NULL,
  event_date = NULL,
  ...
)

# S3 method for class 'event_prices'
print(x, ...)

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

# S3 method for class 'event_prices'
plot(x, ...)

Arguments

x

A data.frame/tibble (or an existing event_prices object, returned unchanged).

...

Reserved for future methods.

time

Name of the time column (character). If NULL, the first of time, date, timestamp, datetime, t (case-insensitive) is used.

price

Name of the price/probability column (character). If NULL, the first of q, price, p, q_t, value (case-insensitive) is used. Ignored when bid and ask are given.

bid, ask

Optional names of bid and ask columns; if both are given, the mid quote is used as price.

scale

Numeric, divisor applied to the raw price first (use scale = 100 for percent quotes; default 1).

discount, book, method

Passed to q_from_price().

clip

Numeric length-2, clipping bounds applied to q before the log-odds transform (default c(0.01, 0.99), see ec_default_params()).

market_id

Optional character label of the market.

event_date

Optional Date/POSIXct of the scheduled event (resolution) date.

object

An event_prices object (for summary()).

Value

An event_prices tibble; see Details.

Details

The constructor follows a strict flag, don't drop convention: raw values are kept in q_raw, data problems are recorded in flag columns, and nothing is silently deleted. The only structural interventions are sorting by time and removing duplicated timestamps (keeping the first occurrence, with a warning), because increments are undefined otherwise.

Columns of the returned object:

time

Date or POSIXct timestamp (sorted, unique).

q_raw

the raw input price after rescaling by scale.

q

the normalized probability, q_from_price(q_raw, ...).

flag_na

TRUE where q is missing.

flag_clip

TRUE where q falls outside the clipping bounds and will be clipped before the log-odds transform in event_clock().

The clipping bounds, market id, and event date are stored as attributes (clip, market_id, event_date) and are picked up by event_clock(), event_clock_path(), and the plotting functions.

Timezones. A POSIXct time column is kept in the timezone it carries; character timestamps are parsed as UTC. Window bounds passed as Date to event_clock() and friends are interpreted in the series' timezone (with to covering the full day), so mixing Date bounds with a non-UTC intraday series is safe.

Functions

  • print(event_prices): Print method; shows market, range, and flags.

  • summary(event_prices): Summary method; returns a one-row tibble including the full-sample event-clock estimate.

  • plot(event_prices): Plot method; dispatches to plot_q().

Examples

data(brexit2016)
ep <- as_event_prices(brexit2016,
  time = "date", price = "q_leave",
  market_id = "Brexit: Leave", event_date = as.Date("2016-06-23")
)
ep
#> -- Event prices: Brexit: Leave
#> 119 observations, 2016-02-26 to 2016-06-23
#> Scheduled event: 2016-06-23
#> # A tibble: 119 × 5
#>    time       q_raw     q flag_na flag_clip
#>    <date>     <dbl> <dbl> <lgl>   <lgl>    
#>  1 2016-02-26 0.312 0.312 FALSE   FALSE    
#>  2 2016-02-27 0.315 0.315 FALSE   FALSE    
#>  3 2016-02-28 0.307 0.307 FALSE   FALSE    
#>  4 2016-02-29 0.307 0.307 FALSE   FALSE    
#>  5 2016-03-01 0.305 0.305 FALSE   FALSE    
#>  6 2016-03-02 0.296 0.296 FALSE   FALSE    
#>  7 2016-03-03 0.287 0.287 FALSE   FALSE    
#>  8 2016-03-04 0.262 0.262 FALSE   FALSE    
#>  9 2016-03-05 0.276 0.276 FALSE   FALSE    
#> 10 2016-03-06 0.279 0.279 FALSE   FALSE    
#> # ℹ 109 more rows
summary(ep)
#> # A tibble: 1 × 11
#>   market_id       n start      end        q_start q_end q_min q_max  n_na n_clip
#>   <chr>       <int> <date>     <date>       <dbl> <dbl> <dbl> <dbl> <int>  <int>
#> 1 Brexit: Le…   119 2016-02-26 2016-06-23   0.312  0.23  0.17   0.4     0      0
#> # ℹ 1 more variable: A_full <dbl>