This package provides a framework for modelling and inferencing attachment mechanisms of temporal complex networks. The two main functionalities of the package are:

• Estimating preferential attachment (PA) and fitness when the growth process can be observed. This is when one can observe the temporal network at at least two time-steps. For estimating the PA function in isolation, we implement Jeong's method, the corrected Newman's method and the PAFit method. For jointly estimating the PA function and node fitnesses, we implement the PAFit method.
• Estimating PA from only one snapshot of the temporal network. We implement the PAFit-oneshot method for estimating the PA function from one single snapshot.

In addition, the package can quantify the remaining uncertainties by providing confidence intervals for the estimated results. We also provide flexible methods to generate a wide range of temporal networks based on PA and fitness.

The release version of the package is hosted on CRAN and can be installed in the usual way:

install.packages("PAFit")

This dev version on GitHub can be installed as follows:

require(devtools)
install_github("thongphamthe/PAFit@devel")

To get started, load the package

library("PAFit")

then work through the tutorial (link to the current CRAN version):

• Tutorial

A version of this tutorial is published in Journal of Statistical Software:

• JSS paper

The PAFit-oneshot method is described in this paper.

Please refer to the current version on CRAN:

• Manual

You can view the html version, which has a better layout but renders mathematical symbols worse than the pdf version, if you use Rstudio

Please refer to the current version on CRAN:

• NEWS

Please refer to the citation information file (link to the current CRAN version):

• Citation Info

GPL-3

• If you have any suggestions or find bugs, please use the github issue tracker
• Feel free to submit pull requests