estimate_from_combined_observations.RdThis function allows for combining two different incidence time series,
see Details .
The two timeseries can represent events that are differently delayed from the original infection events.
The two data sources must not have any overlap in the events recorded.
The function can account for the one of the two types of events to require
the future observation of the other type of event.
For instance, one type can be events of symptom onset, and the other be case confirmation.
Typically, the recording of a symptom onset event will require a future case confirmation.
If so, the partial_observation_requires_full_observation flag should be set to TRUE.
estimate_from_combined_observations( partially_delayed_incidence, fully_delayed_incidence, smoothing_method = "LOESS", deconvolution_method = "Richardson-Lucy delay distribution", estimation_method = "EpiEstim sliding window", delay_until_partial, delay_until_final_report, partial_observation_requires_full_observation = TRUE, ref_date = NULL, time_step = "day", output_Re_only = TRUE, ... )
| partially_delayed_incidence | An object containing incidence data through time. It can be:
|
|---|---|
| fully_delayed_incidence | An object containing incidence data through time. It can be:
|
| smoothing_method | string. Method used to smooth the original incidence data. Available options are:
|
| deconvolution_method | string. Method used to infer timings of infection events from the original incidence data (aka deconvolution step). Available options are:
|
| estimation_method | string. Method used to estimate reproductive number values through time from the reconstructed infection timings. Available options are:
|
| delay_until_partial | Single delay or list of delays. Each delay can be one of:
|
| delay_until_final_report | Single delay or list of delays. Each delay can be one of:
|
| partial_observation_requires_full_observation | boolean
Set to |
| ref_date | Date. Optional. Date of the first data entry in |
| time_step | string. Time between two consecutive incidence datapoints.
"day", "2 days", "week", "year"... (see |
| output_Re_only | boolean. Should the output only contain Re estimates? (as opposed to containing results for each intermediate step) |
| ... | Arguments passed on to
|
Effective reproductive estimates through time.
If output_Re_only is FALSE, then transformations made
on the input observations during calculations are output as well.
With this function, one can specify two types of delayed observations of
infection events (in the same epidemic). The two incidence records are
passed with the partially_delayed_incidence and fully_delayed_incidence.
These two types of delayed observations must not overlap with one another:
a particular infection event should not be recorded in both time series.
If the two sets of observations are completely independent from one another,
meaning that they represents two different ways infection events
can be observed, with two different delays
then set partial_observation_requires_full_observation to FALSE.
Note that a particular infection events should NOT be recorded twice:
it cannot be recorded both in partially_delayed_incidence and in fully_delayed_incidence.
An alternative use-case is when the two sets of observations are not independent
from one another. For instance, if to record a "partially-delayed" event,
one had to wait to record it as a "fully-delayed" event first.
A typical example of this occurs when recording symptom onset events:
in most cases, you must first wait until a case is confirmed via a positive test result
to learn about the symptom onset event (assuming the case was symptomatic in the first place).
But you typically do not have the date of onset of symptoms
for all cases confirmed (even assumed they were all symptomatic cases).
In such a case, we set the partial_observation_requires_full_observation flag
to TRUE and we call the incidence constructed from events of
symptom onset partially_delayed_incidence and
the incidence constructed from case confirmation events
fully_delayed_incidence.
The delay from infection to symptom onset events is
specified with the delay_until_partial argument.
The delay from symptom onset to positive test in this example is
specified with the delay_until_final_report argument.
Note that, for a particular patient,
if the date of onset of symptom is known, the patient must not be counted again
in the incidence of case confirmation.
Otherwise, the infection event would have been counted twice.
shape_incubation = 3.2 scale_incubation = 1.3 delay_incubation <- list(name="gamma", shape = shape_incubation, scale = scale_incubation) shape_onset_to_report = 2.7 scale_onset_to_report = 1.6 delay_onset_to_report <- list(name="gamma", shape = shape_onset_to_report, scale = scale_onset_to_report) ## Basic usage of estimate_from_combined_observations Re_estimate_1 <- estimate_from_combined_observations( partially_delayed_incidence = HK_incidence_data$onset_incidence, fully_delayed_incidence = HK_incidence_data$report_incidence, partial_observation_requires_full_observation = TRUE, delay_until_partial = delay_incubation, delay_until_final_report = delay_onset_to_report ) ## Advanced usage of estimate_from_combined_observations # Getting a more verbose result. Adding a date column and returning intermediate # results as well as the Re estimate. Re_estimate_2 <- estimate_from_combined_observations( partially_delayed_incidence = HK_incidence_data$onset_incidence, fully_delayed_incidence = HK_incidence_data$report_incidence, partial_observation_requires_full_observation = TRUE, delay_until_partial = delay_incubation, delay_until_final_report = delay_onset_to_report, ref_date = HK_incidence_data$date[1], output_Re_only = FALSE ) # Incorporating prior knowledge over Re. Here, Re is assumed constant over a time # frame of one week, with a prior mean of 1.25. Re_estimate_3 <- estimate_from_combined_observations( partially_delayed_incidence = HK_incidence_data$onset_incidence, fully_delayed_incidence = HK_incidence_data$report_incidence, partial_observation_requires_full_observation = TRUE, delay_until_partial = delay_incubation, delay_until_final_report = delay_onset_to_report, estimation_method = 'EpiEstim piecewise constant', interval_length = 7, mean_Re_prior = 1.25 ) # Incorporating prior knowledge over the disease. Here, the mean of the serial # interval is assumed to be 5 days, and the standard deviation is assumed to be # 2.5 days. Re_estimate_4 <- estimate_from_combined_observations( partially_delayed_incidence = HK_incidence_data$onset_incidence, fully_delayed_incidence = HK_incidence_data$report_incidence, partial_observation_requires_full_observation = TRUE, delay_until_partial = delay_incubation, delay_until_final_report = delay_onset_to_report, mean_serial_interval = 5, std_serial_interval = 2.5 )