detect_changepoint: Detect change in temporal trend
In reconhub/epichange: Detect Changes in Temporal Trends

Description Usage Arguments Author(s) See Also

This function looks for the 'optimal' number of recent days to exclude from the temporal trend fitting in asmodee. The procedure selects the value of k which maximises the sum if i) the number of non-outliers in training set (before the last k points) ii) the number of outliers in the last k points. Note that for each value of k investigated, model selection is performed as described in ?asmodee.

detect_changepoint(
  data,
  models,
  date_index,
  alpha = 0.05,
  max_k = 7,
  method = trendeval::evaluate_resampling,
  include_warnings = FALSE,
  ...
)

`data`	A `data.frame` or a `tibble` containing the response and explanatory variables used in the `models`.
`models`	A list of `trending_model()` objects, generated by `lm_model`, `glm_model`, `glm_nb_model`, `brms_model` and similar functions (see `?trending::trending_model()`) for details.
`date_index`	The name of a variable corresponding to time, quoted or not.
`alpha`	The alpha threshold to be used for the prediction interval calculation; defaults to 0.05, i.e. 95% prediction intervals are calculated.
`max_k`	An `integer` indicating the maximum number of recent data points to be excluded from the trend fitting procedure. By default, ASMODEE will look for a changepoint within this recent time period, after which data no longer fit the previous trend. Larger values will require more computation from the method. Only used if `fixed_k` is `NULL`.
`method`	A function used to evaluate model fit. Current choices are `evaluate_aic` (default) and `evaluate_resampling`. `evaluate_aic` uses Akaike's Information Criterion instead, which is faster but possibly less good a selecting models with the best predictive power. `evaluate_resampling` uses cross-validation and, by default, RMSE to assess model fit.
`include_warnings`	Include results in output that triggered warnings but not errors. Defaults to `FALSE`.
`...`	Further arguments passed to `method`.