- add option to use instantaneous R rather than the default case R in project

- harmonise si definition with EpiEstim.

- now using 3rd edition of testthat for checks.

now using the new default random number generator in tests

updated tests to pass check with R 4.0.2; this includes re-generation of most reference objects, which were manually inspected and validated against old references

new function

`merge_projections`

permits to join different sets of simulations, adapting the respective time windows accordinglynew function

`merge_add_projections`

permits to add simulated case incidence from different sets of simulations, adapting the respective time windows accordingly; if objects contain different numbers of simulations, the shortest ones are recycled as needed; also implemented as the operator`+`

new function

`summary`

will summarise`projections`

objects by day, using a range of pre-defined statistics (mean, sd, min, max, and user-defined quantiles)

bug fix for the use of time-varying R in

`project`

bug fix for

`as.data.frame(..., long = TRUE)`

more consistent handling of inputs for the serial interval; if provided as a vector,

`si`

now starts at a delay of 1 day, rather than 0 and assuming the first entry is 0; for`distcrete`

inputs, the mass for`si$d(0)`

is now ignored, and the rest of the distribution is rescaled accorindly to ensure the PMF sums to 1(

*documentation*) more realistic serial interval distribution used for the Ebola examples(

*internal*) the force of infection is now calculated by the new internal function`compute_force_infection`

(

*testing*) revised sets of tests relying less on comparison to references, but testing for meaningful output properties instead(

*testing*) 100% coverage

`project()`

will can now take single dates or single projections as inputs. (@acori, @zkamvar, #18).

`project`

can now use time-varying R by specifying`time_change`

, a vector of dates at which R changes, and providing a`vector`

or a`list`

of values for`R`

instead of a vector, in which case it needs to have`length(time_change) + 1`

components, each of which is a vector of R values.

- A bug in
`project()`

where R was being resampled recursively was fixed (#11, @jarvisc1; #12, @zkamvar)

`get_dates()`

now inherits the generic`get_dates()`

from incidence

This is a big release! Plenty of new features have been added, including:

`projections`

can now be subsetted like matrices using`x[i,j]`

, or using the function`subset`

new function

`cumulate`

to compute cumulative incidence for`projections`

objects, akin to the similar function in the`incidence`

packagemuch improved graphics;

`plot`

now call upons`add_projections`

, which implements many options for plotting projections, including quantiles lines, ribbon, and boxplots;`add_projections`

can also be used to add such plots to an existing`incidence`

plot`build_projections`

can be used to build a`projections`

object from an input matrix and optional dates

`project`

can now use a Negative Binomial model as an alternative to the original Poisson model to account for over-dispersion in new cases / super-spreading.

Fixed the calls to `sample`

, which has a rather irrational
behaviour when passing a first argument of length 1 (i.e. not
bootstrapping from the vector).

First release of the package!

`project`

: a function generating projections from an existing*incidence*object, a serial interval distribution, and a set of plausible reproduction numbers (\(R\)); returns a`projections`

object.`plot`

/`print`

: plotting and printing methods for`projections`

objects.`get_dates`

: accessors for`projections`

objects.`as.data.frame`

: conversion from`projections`

objects to`data.frame`

.