18 Appendix

Here you can find some additional materials that provide more details about topics that we address only briefly in the lectures. They are organized in the same sequence as the module sections.

18.1 Data requirements

18.1.1 Example egocentric partner module

During the early 2000’s the Demographic and Health Surveys (DHS) HIV/AIDS indicators program collected egocentric sexual network data from representative samples of the population in over 25 countries around the world.

An example partnership module from the DHS questionnaire is embedded below.

This is a remarkable resource for cross-country analyses. Some countries also had surveys at multiple time points, which, in theory, would allow for estimation of changes in behavioral patterns.

There were, however, some issues with the module design that limit the usefulness of the data.

18.1.1.1 Module variability

There was some variability in the module structure by country over time. Some collected data on the last 3 sexual partners in the last year, some on the last 2 only. Some had information on the date of last sex, some did not. In addition, some countries only included women in their surveys. This complicates the analysis of changes over time and differences between countries.

18.1.1.2 Failure to assess partnership status

There was a key omission in all of the versions of the DHS partner module: they did not collect information on the current status of relationships – whether the partnership had ended or was still active.

The result is that one needs to make assumptions about the active status of a partnership in order to estimate the mean age of active ties. And that is the key parameter needed to parameterize the STERGM for EpiModel.

This omission also influences the estimation of concurrency in these populations. See the paper below for more details:

Morris M, Vu L, Leslie-Cook A, Akom E, Stephen A, Sherard D. Comparing Estimates of Multiple and Concurrent Partnerships Across Population Based Surveys: Implications for Combination HIV Prevention. AIDS Behav. 2014 Apr;18(4):783-790. https://doi.org/10.1007%2Fs10461-013-0618-6

18.1.2 Estimating mean duration of active ties

Details on the relationship between dissolution dynamics, sampling and estimating relationship duration. Includes a link to an RShiny App for building intuition.

18.2 Model specification

18.2.1 Understanding degeneracy

Slideset with more details on how these degenerate models behave.