Event Sequence Analysis

Event sequence analysis studies trajectories as ordered changes.

Instead of asking what state is a person in at each time point?, it asks:

• What events happened?
• In what order did they happen?
• Did some events happen at the same time?
• Are some event patterns more common in one group than another?

This makes event sequence analysis especially useful when the main object of interest is the ordering of life events, career events, educational transitions, medical episodes, customer actions, or other discrete changes.

A First Intuition

Suppose we observe a person from age 18 to 30.

A state sequence may look like this:

Age	18	19	20	21	22	23
State	Education	Education	Education	Full-time work	Full-time work	Married

This tells us what state the person occupies at each age.

The corresponding event sequence may look like this:

(Start education) -> (Start full-time work) -> (Marry)

This tells us what changed, and in what order.

Both representations describe the same trajectory, but they highlight different aspects of it. A state sequence is often better for studying duration, such as how long someone stays in education or employment. An event sequence is often better for studying ordering, such as whether people usually leave education before entering work, or whether marriage tends to occur before or after childbirth.

State Sequences and Event Sequences

The key difference is the meaning of a position.

In a state sequence, position usually means time. For example, position 5 may mean the fifth month, fifth year, or fifth observation wave.

In an event sequence, position means event order. Position 5 means the fifth event, not necessarily the fifth year. If the timing of events matters, timestamps need to be stored explicitly.

This distinction is central to event sequence analysis. Events do not last; they occur at particular points and may cause changes in states. Several events can also occur at the same time. For example, a person may leave the parental home and move in with a partner in the same year:

(Leaving home, Union)

Here, the comma means that the two events are simultaneous. The arrow means that one transition comes after another:

(Leaving home, Union) -> (Marriage) -> (Childbirth)

This notation makes event sequence analysis well suited for studying ordered and possibly simultaneous changes.

What Can We Learn from Event Sequences?

Event sequence analysis is useful when the research question is about patterns of change rather than only time spent in states.

Typical questions include:

• What are the most common event patterns?
• Which events tend to occur together?
• Which event usually follows a given event or subsequence?
• Which subsequences distinguish one group from another?
• Do different cohorts, genders, countries, or organizations follow different event patterns?
• Does a theoretically important pattern appear in each sequence?

For example, in life course research, we may ask whether people tend to leave home before entering a union, whether childbirth follows marriage, or whether part-time work tends to appear after childbirth. In career research, we may ask whether education is followed by full-time employment, whether unemployment interrupts work trajectories, or whether promotion tends to occur after job changes.

The Basic Workflow

A typical event-sequence workflow in Sequenzo has four steps.

1. Build event sequences

You can create event sequences from time-stamped event data:

from sequenzo.event_sequences import EventSequenceData

eseq = EventSequenceData.from_tse(data=tse_df)

The input data should usually contain one row per event, with at least three columns:

id	timestamp	event
1	18	EnterEducation
1	22	StartFullTimeWork
1	27	Marry
2	19	EnterEducation
2	21	StartFullTimeWork

You can also create event sequences from existing state sequences:

eseq = EventSequenceData.from_state_sequences(seqdata)

This is useful when your data are originally stored as state sequences, but you want to analyze the changes behind those states.

2. Find frequent subsequences

Once the event sequences are created, you can search for common event patterns:

from sequenzo.event_sequences import find_frequent_subsequences

fsubseq = find_frequent_subsequences(
    eseq,
    min_support_ratio=0.05
)

If timestamps are available, search_constraint can restrict which subsequences count as matches, for example by requiring events to occur within a maximum time span.

A frequent subsequence is an ordered event pattern that appears in many sequences. For example:

(Education) -> (Full-time work)
(Leaving home, Union) -> (Childbirth)

The first pattern says that education is followed by full-time work. The second says that leaving home and entering a union happen together, followed later by childbirth.

3. Compare groups

You can then ask which subsequences distinguish groups:

from sequenzo.event_sequences import compare_groups

discr = compare_groups(
    fsubseq,
    group_labels=df["group"],
    pvalue_threshold=0.05
)

This helps answer questions such as:

Which event patterns are more common among women than men?
Which patterns are more common in younger cohorts than older cohorts?
Which career sequences distinguish two occupational groups?

4. Visualize the results

Sequenzo provides several plotting functions for communicating event-sequence results:

from sequenzo.event_sequences import plot_subsequence_frequencies

plot_subsequence_frequencies(fsubseq)

You can also visualize group contrasts, event-order structures, and event dynamics.

Main Functions

Task	Sequenzo function
Create event sequences from time-stamped event data	EventSequenceData.from_tse()
Create event sequences from state sequences	EventSequenceData.from_state_sequences()
Find frequent subsequences	find_frequent_subsequences()
Count subsequence occurrences	count_subsequence_occurrences()
Compare groups	compare_groups()
Check whether a target pattern appears	check_event_subsequence_containment()
Convert event sequences back to TSE format	convert_event_sequences_to_tse()
Summarize adjacent event-order movements	compute_event_transition_matrix()
Plot frequent subsequences	plot_subsequence_frequencies()
Plot group contrasts	plot_subsequence_group_contrasts()
Plot event parallel coordinates	plot_event_parallel_coordinates()
Plot event dynamics	plot_event_dynamics()

For TraMineR Users

This module mirrors the practical logic of TraMineR’s event-sequence tools, while using Python-style names.

TraMineR	Sequenzo
seqecreate()	EventSequenceData.from_tse() / EventSequenceData.from_state_sequences()
seqefsub()	find_frequent_subsequences()
seqeapplysub()	count_subsequence_occurrences()
seqecmpgroup()	compare_groups()
seqecontain()	check_event_subsequence_containment()
seqe2tse()	convert_event_sequences_to_tse()
seqetm()	compute_event_transition_matrix()
seqelist	EventSequenceList
eseq	EventSequence

Relationship with Other Sequenzo Modules

Use sequenzo.event_sequences when your question is about patterns of events:

Which events happen, in what order, and for whom?

Use sequenzo.with_event_history_analysis when your question is about event-history or sequence-history modeling:

What is the risk or hazard of a transition, given previous history?

In short:

Patterns of events -> event_sequences
Modeling transitions or hazards -> with_event_history_analysis

Reference

Ritschard, G., Bürgin, R., & Studer, M. (2013). Exploratory Mining of Life Event Histories. In J. J. McArdle & G. Ritschard (Eds.), Contemporary Issues in Exploratory Data Mining in the Behavioral Sciences (pp. 221–253). Routledge.