Relational Hyperevent Outcome Models

[Megan Chan]

Hi all,

I've seen really interesting work on relational hyperevent outcome models (here and here), and I was curious about the most straightforward approach to model them. Can you simply use the eventnet package to generate statistics and modify the R code to model event outcomes instead of event occurrences?

Thanks!

[juergen lerner]

Dear Megan,

Yes - to specify and estimate relational hyperevent outcome models
(RHOM) you can use eventnet to compute hyperedge statistics and then
use R (or other statistical software) to model the outcome variable
dependent on the statistics.

When specifying the observations in eventnet, you typically condition
on the observed hyperedge. (For undirected hyperedges, check
"condition on node or UHE"; for directed hyperedges check "condition
on source" and check "condition on target".) This is because in RHOM
you don't seek to explain who participates in the event, but you
condition on an observed group of event participants and want to
explain the outcome of that event.

The outcome is usually given as the event weight (column WEIGHT in the
eventnet output file) and you can model it, for example, by standard
regression models that are appropriate for the type of variable. For
example, linear regression for numeric outcome or logistic regression
for binary outcome.

There is no eventnet tutorial on RHOM, yet. However I can point you to
detailed material by Martin Koch who used RHOM to explain the success
of creative teams:

https://github.com/creamartin/master_css

(That's also linked from the eventnet github page under "third-party
material".)

I hope this helps.

Best wishes
Juergen

> --
> You received this message because you are subscribed to the Google Groups "eventnet-users" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to eventnet-user...@googlegroups.com.
> To view this discussion visit https://groups.google.com/d/msgid/eventnet-users/df6be244-9c54-4a9d-9560-d8aed5c81041n%40googlegroups.com.

[Megan Chan]

Thank you so much! This is exactly what I'm looking for.