Phase 2 of user interface: Type conversion functions #210
Description
We want to make the following helper functions:
as_reporting_triangle()
as_data_frame()
Questions remaining:
- Do we need to create a reporting triangle class? I would think yes in order to save the matrices and associated metadata if we have different strata e.g. age groups, locations, etc.
- Could also create a dataframe with metadata on strata remaining as columns?
- Do we want a method to convert linelist data as well, and how do we detect this, given it is also a dataframe?
- Does this require the user to specify a max delay or can the maximum delay be inferred from the dataframe (e.g. the max difference with a count greater than 0)?
Context
Type Conversion Functions
Add utility functions for data format conversion:
# Convert dataframe to reporting triangle (S3 generic)
as_reporting_triangle <- function(data, ...) UseMethod("as_reporting_triangle")
# S3 method for data.frame
as_reporting_triangle.data.frame <- function(
data,
reference_date,
report_date,
count,
strata = NULL,
max_delay = 30, # Now here instead of main function
...
) {
# Creates a reporting triangle with attributes:
# - reference_dates: vector of dates matching triangle rows
# - strata: strata values if applicable
# - max_delay: maximum delay value used
}
# S3 method for epinowcast data
as_reporting_triangle.enw_data <- function(
data,
max_delay = 30,
...
) { ... }
# Convert reporting triangle/nowcast to dataframe with dates
as_data_frame <- function(
triangle_or_nowcast,
reference_dates = NULL, # Vector of dates corresponding to triangle rows
strata = NULL
) {
# If reference_dates is NULL, try to get from attributes
if (is.null(reference_dates)) {
reference_dates <- attr(triangle_or_nowcast, "reference_dates")
if (is.null(reference_dates)) {
stop("Reference dates must be provided or available in attributes")
}
}
# Map matrix rows back to dates and return data.frame
}