SARML has been replaced by sar.SARTML has been replaced by sart.CDnetNPL has been replaced by cdnet.simSARnet has been replaced by
simsar.simTobitnet has been replaced by
simsart.simCDnet has been replaced by
simcdnet.netformation has been replaced by
homophily.It is now possible to estimate the SART model under rational expectations. In the previous version, the SART model is only available under complete information.
This version follows the major revision of the paper in September 2022. - The count data model includes a more flexible specification. Especially, it is possible to assume that the cut points are not equally spaced for large values of the dependent variable. - I also implement a network formation model with degree heterogeneity as fixed effects (see Yan et al., 2019). - Models under incomplete information are now estimated using LBFGS algorithm of the package RcppNumerical. Thus, the optimization is performed in C++ and is very fast compared to the version 1.0.1.
Note and Warning found in the check for MACOS have been fixed.
R defaulted to C++11 in R 4.0.0, to C++14 in R 4.2.0 and to C++17.
Fixed effect is allowed in the model SAR.
homophily has been changed to homophily.re
for the random effect models.homophily.FE has heen changed to
homophily.fe.homophily.re can be one side or two
sides.homophily.fe can be one side or two
sides.homophily.re.homophily.fe.homophili.data is added to convert data
between directed network models and symmetric network models.cdnet now
allows heterogeneity in peer effects. For example, peer effects can be
estimated separately for groups such as Blacks and Whites, or Girls and
Boys.cdnet now also allows heterogeneity in the cut-points.
The cost function in the utility function is nonparametric and depends
on characteristics X such as gender, race, etc.homophili.fe, elements in the estimate and gradient
vectors are now labelled.cdnet and sart no
longer displays marginal effects.meffects method has been added and can be applied
to the outputs of cdnet and sart.