Workshop

Network Dynamics using sienaBayes

Groningen, September 7, 2018



Basic Introduction

Don't use Internet Explorer as a browser for these pages. It does not reproduce the R scripts in a nice fashion.

This is a workshop about multilevel longitudinal network analysis using the function sienaBayes in package RSienaTest. This is a function that can be used when you have a data set with many 'parallel' networks, called 'groups', all with the same networks and other variables but collected on different groups of actors. The group sizes can be different. The use of sienaBayes is meant for data sets where each group by itself would be too small for an analysis by siena07, and where, like in other statistical multilevel approaches, the analysis of each group can borrow strength from considering it as being similar to the others; formally, from considering the set of groups as a sample from a population of longitudinal network data sets.

The workshop takes place Friday September 7, 2018, from 9.30 to 17.00.



Homework

  1. Install the new version 1.2-13 of RSienaTest on your computer. The links are below.
  2. Find out the number of processes you may use on the computer that you use for this type of work, by the function detectCores as is done, e.g., in the help page of siena07.
  3. If you do not already know about this, learn the basics of how to work with lists in R: create lists, and how to use the functions lapply and sapply. You can find a lot of material about this on the web.
    If you have never used the functions save and save.image, also study how to use them.
  4. To get a first practical idea of Bayesian estimation for network models, study and run the script IntroToBayes.R which you will find below.
  5. Get acquainted with the basics of running sienaBayes, by carrying out the example at the bottom of the Help Page for sienaBayes.
  6. Get further acquainted with the basics of running sienaBayes by following the script RscriptsienaBayes.r which you will find below.
  7. Organize your own data set in a Siena multiple group data set. For working with it in the workshop it will be good that it is not too large, e.g., 10 groups. (Note that this is on the small side for use of sienaBayes.)
  8. Find out for your own data (or the subset that is not too large) how large the value of mult should be to get the autocorrelations less than 0.4, as done also in script RscriptsienaBayes.r.


Materials

The materials all are work in progress.

Software

Texts

Scripts

  • IntroToBayes.R, an introductory script to Bayesian estimation for the simplest network model: the Bernoulli graph, where the network is non-directed and non-longitudinal, all edges are independent and have the same probability p of being present, but the issue still is that p is unknown...
  • RscriptsienaBayes.r, an example script for the function sienaBayes in RSienaTest, using a multi-group data set from Andrea Knecht's data, of which a description of one class is to be found here.
    Modified version (for EUSN 2019), September, 2019.
  • RscriptsienaBayes_2.r, another example script for the function sienaBayes in RSienaTest, with a further multilevel model specification and various further options (added later).
  • BayesPlots.r, a collection of plotting functions for sienaBayesFit objects, produced by sienaBayes.

Introductions to Bayesian statistics

There are many introductions to Bayesian statistics on the web. Some more, others less technical. Various of these have an anti-frequentist bias. These are some others:




Back to the main Siena page