When there is a lag in the reception of data nowcast can be used to correct for this lag, and predict the true underlying values.
Usage
nowcast(
data_aggregated,
offset,
n_week_adjusting,
n_week_training,
date_0,
nowcast_correction_fn = nowcast_correction_fn_negbin_mm,
nowcast_correction_sim_fn = nowcast_correction_sim_neg_bin
)Arguments
- data_aggregated
Aggregated dataset from the function nowcast_aggregate
- offset
offset
- n_week_adjusting
Total number of weeks to correct.
- n_week_training
Number of weeks to train on
- date_0
Date of aggregation.
- nowcast_correction_fn
Correction function. The deafault is nowcast_correction_fn_expanded. Must return the same as this function.
- nowcast_correction_sim_fn
Simmulatoin function. Must return a datatable with the following collumns "n_death", "sim_value", "cut_doe", "ncor" and simmulations for equally many weeks as n_week_adjust.
Details
For more details see the help vignette:
vignette("nowcast", package="attrib")
Examples
data_aggregated <- attrib::data_fake_nowcasting_county_aggregated
n_week_training <- 50
n_week_adjusting <- 1
date_0 <- data_aggregated[order(cut_doe)][nrow(data_aggregated)]$cut_doe + 1
offset = "log(pop)"
nowcast_object <- attrib::nowcast(
data_aggregated,
offset,
n_week_adjusting,
n_week_training,
date_0
)
data_aggregated <- attrib::data_fake_nowcasting_county_aggregated
data_aggregated <- data_aggregated[location_code == "county_nor03"]
n_week_training <- 50
n_week_adjusting <- 1
date_0 <- data_aggregated[order(cut_doe)][nrow(data_aggregated)]$cut_doe + 1
offset = "log(pop)"
nowcast_object <- nowcast(
data_aggregated,
offset,
n_week_adjusting,
n_week_training,
date_0,
nowcast_correction_fn = nowcast_correction_fn_quasipoisson,
nowcast_correction_sim_fn = nowcast_correction_sim_quasipoisson
)