gttrend_adj: Trend adjustment of a Google Trends time series

In johannes97s/ifwtrends: Evaluating Google Trends data for economic forecasts

Trend adjustment of a Google Trends time series

Description

This function can either trend adjust a given time series or create a new time series based on a Google Trends search query and then directly trend adjust it.

Usage

gttrend_adj(
  timeseries = NULL,
  keyword = NA,
  category = NA,
  geo = "DE",
  timeframe = paste("2006-01-01", Sys.Date()),
  method = "moving_avg",
  log.trafo = FALSE
)

Arguments

timeseries	A already created time series to trend adjust. Preferably already as tsibble or tibble. If another data format is given, it will be coerced to a tsibble.
keyword	A vector (chr) of keywords to search for.
category	A vector (numeric) of category numbers to search for.
geo	The region to search in.
timeframe	A time frame to search the queries in consisting of a start date and an end date in YYYY-MM-DD form.
method	As trend adjustment method, one can choose between "moving_avg", "firstdiff", and "comtrend." See Details for more information.
log.trafo	Logical, indicates if value should be transformed to log(value).

Value

Returns a tibble with trend adjusted values and a date column.

For a trend method, there can be choosen between "firstdiff", "moving_avg" and "comtrend". If you choose "firstdiff", first differences with lag = 1 are computed. If you choose the moving average, the time series will be decomposed into its components and the trend will be subtracted from the whole time series (using loess). With "comtrend", there is a polynom of degree 5 with id-fixed effects estimated, which captures the common trend of a sample of categories in Google Trends. The residuals are then used as the adjusted series. For further detail, see Woloszko et al. (2020) and the function est_trend().

Examples

# Trend adjusting a already established series.
series <- trendecon::ts_gtrends("ikea", time = "all")
gttrend_adj(series, log.trafo = TRUE, method = "moving_avg")

# Search for a new series and trend adjust it
gttrend_adj(
  category = 179, timeframe = "2015-01-01 2021-01-01",
  method = "moving_avg", log.trafo = FALSE
)

johannes97s/ifwtrends documentation built on Oct. 9, 2022, 7:01 p.m.