In EcoGRAPH/epidemiar: epidemiar: Create EPIDEMIA Environmentally-Mediated Disease Forecasts

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

Epidemiar Output Dataset - Full Run

The results of run_epidemia() is a named list of the following datasets or objects:

1. summary_data
2. epi_summary
3. modeling_results_data
4. environ_timeseries
5. environ_anomalies
6. params_meta
7. regression_object

summary_data

Early detection and early warning alerts levels for each geographic group. Early detection alerts (ed_alert_count) are alerts that are triggered during the early detection period, which was defined by the argument ed_summary_period, the n most recent weeks of known epidemiology data. Similarly, early warning alerts (ew_alert_count) are alerts in the future forecast estimates. “High” level indicates two or more weeks in this period had incidences greater than the alert threshold, “Medium” means that one week was in alert status, and “Low” means no weeks had alerts (ed_sum_level and ew_level, respectively).

• {groupfield}: The user-given geographic grouping field
• ed_alert_count: Number of alerts triggered in the early detection period
• ed_sum_level: High/Medium/Low depending on the number of alerts, 2+/1/0 respectively
• ew_alert_count: Number of alerts triggered in the early warning period (future forecast period)
• ew_level: High/Medium/Low depending on the number of alerts, 2+/1/0 respectively

epi_summary

Mean disease incidence per geographic group during the early detection period.

• {groupfield}: The user-given geographic grouping field
• mean_epi: The mean disease incidence (or cases, depending on the setting in report_settings$report_value_type) per geographic group summarized over the early detection period

modeling_results_data

These are multiple timeseries values for observed, forecast, and alert thresholds of disease incidence, over the report period, for each geographic unit. These data can be used in creating the individual geographic unit control charts.

• {groupfield}: The user-given geographic grouping field
• obs_date: The last day of the epidemiological week, Date object
• series: "obs" = observed disease incidence, "fc" = modeled/forecast incidence values, "thresh" = event detection threshold values, "ed" = early detection alert (binary), "ew" = early warning alert (binary)
• value: Value of the series for that geographic group for that week
• lab: Labels for the series ("Observed", "Forecast Trend", "Alert Threshold", "Early Detection Alert", "Early Warning Alert")
• upper: Unused
• lower: Unused
• week_epidemiar: ISO/CDC week number, based on user given report_settings$epi_date_type argument
• year_epidemiar: ISO/CDC year, based on user given report_settings$epi_date_type argument

environ_timeseries

These are multiple timeseries for the used environmental variables during the report period for each geographic unit.

• {groupfield}: The user-given geographic grouping field
• {obsfield}: The user-given field for the environmental variable name/ID
• year_epidemiar: ISO/CDC year, based on user given report_settings$epi_date_type argument
• week_epidemiar: ISO/CDC week number, based on user given report_settings$epi_date_type argument
• obs_date: The last day of the epidemiological week (ISO/CDC, by report_settings$epi_date_type), Date object
• val_epidemiar: Value of the environmental variable for that geographic group for that week. Values are a combination of observed, or interpolated (for missing) or extended (future estimated) values.
• reference_method: Method for creating a weekly summary from daily data (e.g. "sum" for rainfall, or "mean" for NDWI)
• data_source: "Observed" or "Imputed". Environment data was either observed, or if it was NA/missing, it was filled in (imputed). For gaps less than 2 weeks, the values are filled in with a persistence method (carry-forward). The recent values are calculated as the average of the past 7 days for 'mean' type variables (as defined in the user's environ_info metadata, e.g. for NDWI, LST), or the past 14 known days for 'sum' type variables (as defined in the user's environ_info metadata, e.g. for precipitation-like measures). For periods longer than 2 weeks, daily values were imputed using a progressive blend of the recent values (as above) with the climatology/historical averages for that week of the year (from environ_ref_data).
• ref_value: From env_ref_data.
• ref_*: Fields from env_ref_data that begin with ref_ have been propagating through to here. (Potentially useful for plotting, for example.)

environ_anomalies

These data are the recent (during the early detection period) differences (anomalies) of the environmental variable values from the climatology/reference mean.

• {groupfield}: The user-given geographic grouping field
• {obsfield}: The user-given field for the environmental variable name/ID
• anom_ed_mean: The mean of the anomalies per environmental variable per geographic group summarized during the early detection period. The anomalies here are calculated as the difference from the observed value to the historical mean for that week of the year. (Not to be confused with environmental anomalies option in modeling.)

params_meta

This lists dates, settings, and parameters that run_epidemiar() was called with and defaults that were used if the user did not set values for those parameters.

regression_object

This is the regression object from the general additive model (mgvc::bam()) regression. This is generally only for additional statistical investigation of the model, and is usually not saved (large object).

Epidemiar Output Dataset - Model Only Run

The results of run_epidemiar(..., report_settings$model_run = TRUE) is a cached model: the regression object plus some metadata information about what was used to generate the model. Once a model has been generated, it can be fed back into run_epidemiar(..., report_settings$model_cached = {cached model object}) for faster predictions rather than regenerating the model on each run. Determining the balance on how old of a model is still useful is heavily dependent on the specific dataset.

1. model_obj
2. model_info

model_obj

The output regression object from the mgcv::bam() general additive model regression call, or a list of models per cluster from clusterapply depending on the model settings.

model_info

A list of dates, settings, and relevant parameters that run_epidemiar() was called with. Very similar to params_meta of a full run.

EcoGRAPH/epidemiar documentation built on Nov. 13, 2020, 5:31 p.m.