Implements methods for anticipating the emergence and eradication of infectious diseases from surveillance time series. Also provides support for computational experiments testing the performance of such methods.
spaero
is being developed to support the work of researchers working
on Project AERO. spaero
is
publicly available in part to disseminate and increase the
reproducibility of the project's research and in part because it may
be useful to others.
The stable version of spaero
that is on CRAN may be installed from
within R:
install.packages("spaero")
Alternatively, the development version that is on GitHub may be
installed using the devtools
package:
devtools::install_github("e3bo/spaero")
After installation, a listing of the available documentation can be obtained:
help(package = "spaero")
Please use GitHub issues or pull requests to let us know about any bugs you find.
Feature requests are also welcome but be forewarned that we have limited time to work on items not directly supporting the project's research.
More details about contribution may be found in the contribution guidelines.
GNU GPL version 2 or later
Add vaccination reaction to simulator. A vaccination rate of zero remains the default parameter setting.
Add backward-looking window option for get_stats.
Avoid errors when input time series is constant and return a missing value instead.
Add transmission argument to create_simulator to allow for frequency-dependent transmission. Density-dependent transmission remains the default model.
Add vector of first difference of the variance vector produced by get_stats. This change makes it easier to use the convexity of the variance time series as an early warning signal. The name of the vector in the stats list is variance_first_diff. Note that this change makes the abbreviation stats$var ambiguous. Code using that abbreviation to obtain the vector of variance estimates should substitute in stats$variance.
To the output of get_stats(), add list taus containing Kendall's correlation coefficient of the elements of each time series in the stats list in the output with time.
Ensure variance and kurtosis esimates are non-negative. When using local linear for estimating statistics, it was possible in previous versions for negative values to occur.
Correct autocorrelation calculation. The previous version divided the autocovariance by the variance at the most recent time point. The current version divides by the geometric mean of the variance at each of the two time points, matching standard practice. The formula in the vignette for the autocorrelation has been changed accordingly.
Clean up sloppy usage of the term statistic in the documentation.