The events Package

The aim of events is to make life a bit easier for people who analyse event data: the kind of thing that KEDS/TABARI generates as output (see e.g. here).
There's nothing fancy in the package, just a hopefully logical interface to all the data massaging we do to event data before any actual analysis.

The philosophy of the package is that you will want to read in raw event data and then apply a sequence of filters and aggregators for actors and event types, each of which will result in an new event data set. The final step is an application of make_dyads to make a set of named temporally regular directed dyad time series suitable for time series analysis. All rather unix-ish, if you care for that sort of thing.

There's a fairly complete walkthrough in the package vignette. You should probably read that.


The package is ultimately intended to unify the existing software, e.g. the packages currently linked from the PSU event data pages.

The unification is certainly not complete, but we're getting there. In particular, extremely large data sets are probably going to be rather unwieldy in the current version.

The sections below provide a quick compare and contrast to the software available from PSU that works with event data output (not Factiva stuff or actor dictionaries).


This functionality is implemented as a standalone function scrub_keds and as an option in the event data reading function read_keds.


This is also a standalone function one_a_day and available as an option in the event data reading function read_keds.


Most, but not all, of the functionality of KEDS_Count is implemented.
Some differences are noted below:


Unlike Aggregator, events does not have a customizable aggregation period, but it is a lot more flexible about what the aggregation function is (see above).


Since I can't figure out what this program does, there are probably no functions to replicate it.

High-Volume Processing Suite

All this stuff is currently out of reach, but there are plans for accessing databases as sources of event data. That'll be how we scale.