In forvis/forvision: Tools for forecast visualisation and evaluation

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

In time series table schema each observation is stored in a table as a separate record (line):

| Field name (column name)| Description| Examples | | --- | --- | ---- | | series_id | Time series identifier - a unique name that identifies a time series | “Y1” | |timestamp|Any representation of the period to which the observation relates. We recommend the use of the ISO 8601 standard|"1997" in case of yearly data, "1997-01-20" in case of daily data, "1997-11" in case of monthly data, "1997-W03" in case of weekly data, "2018-Q2" in case of quarterly data| |value|The value observed |1000|

*the key (the unique value that should not duplicated) for this table schema is series_id, timestamp. In other words, we cannot have two (or more) records in a table relating to the same time series and the same period of observation (timestamp)

We may have additional fields (columns) in this table or additional table specifying the features of time series. However, the above schema includes the fields that are necessary for further processing of time series data containing many time series. Here we do not impose restrictions on the types of the data.

If some observation is missing, the corresponding table line can be omitted or corresponding value can be denoted as ‘NA’. Observation can also contain censored data, etc., which can also be represented by additional agreements, but here these rules are not specified.

Example:

library(forvision)
head(m3_yearly_ts, 10)

forvis/forvision documentation built on April 30, 2020, 3:28 a.m.