search

leifeld / tnam

Temporal Network Autocorrelation Models (TNAM)
11 stars 4 forks source link

database model #6

Closed souzacharles closed 1 year ago

souzacharles commented 1 year ago

Hello, I intend to apply the tnam package in an academic research. More precisely, in a study on international trade networks (comtrade data). Would you like to know how the database should be built? For example data.frame or list? Do you have any models? Thanks.

leifeld commented 1 year ago

Hi. Please make sure you understand that the way tnam estimates the model is through maximum pseudolikelihood (MPLE), which is likely biased to some extent (depending on the application). If there are no temporal dependencies, consider the new BANAM package (with Bayesian estimation). If you require temporal dependence, consider the M-STAR model (the multiparametric spatio-temporal autoregressive model). A Stata implementation is available here, but maybe R implementations exist, too. I suppose some people write this up in BUGS or Stan to be flexible.

If you still want to use tnam, check out the help pages for tnam and tnamdata for some examples.