get_group_differences()

get_group_differences() compares two predefined sequence groups at the whole-sequence level using LRT and/or BIC.

Function Usage

get_group_differences(
    seqdata,
    seqdata2=None,
    group=None,
    set_var=None,
    s=100,
    seed=36963,
    stat="all",
    squared="LRTonly",
    weighted=True,
    opt=None,
    BFopt=None,
    method="OM",
    **kwargs
)

TraMineR Parameter Mapping

• seqdata, seqdata2, group, set_var -> TraMineRextras seqCompare data/group inputs
• s, seed, stat, squared, weighted, opt, BFopt -> seqCompare control arguments

Entry Parameters

Parameter	Required	Type	Description
seqdata	✓	SequenceData / DataFrame / list	Main input sequences. Can be one dataset or a list of datasets.
seqdata2	✗	SequenceData / DataFrame / list / None	Optional second sequence input. Required if group is not provided.
group	✗	array-like / None	Group labels used to split one seqdata into two groups. Must have exactly 2 valid levels.
set_var	✗	array-like / None	Optional strata variable. If provided, one result is computed per stratum.
s	✗	int	Resampling size used in the comparison engine. Use 0 for full comparison without sampling.
seed	✗	int	Random seed for reproducible resampling.
stat	✗	str	Which statistics to return: "LRT", "BIC", or "all".
squared	✗	bool / str	If string, only "LRTonly" is valid.
weighted	✗	bool / str	Use weights (True/False) or "by.group" for group-wise normalization.
opt	✗	int / None	Internal sampling option. Leave as None unless you need low-level control.
BFopt	✗	int / None	Controls Bayes-factor output style in multi-sample settings.
method	✗	str	Distance method passed to get_distance_matrix() (for example "OM").
**kwargs	✗	any	Extra distance-method arguments forwarded to get_distance_matrix().

What It Returns

A NumPy array of numeric results (column layout follows the internal table built before return):

• If stat="LRT": LRT, p-value
• If stat="BIC": Delta BIC, Bayes-factor column(s)
• If stat="all": LRT columns plus BIC columns

When resampling produces multiple subsamples (multsple), extra Bayes-factor columns may appear depending on BFopt (see source). When set_var is used, rows correspond to strata.

Example

from sequenzo.group_comparison import get_group_differences

result = get_group_differences(
    seqdata=seqdata,
    group=df["group"],
    stat="all",
    method="OM",
    indel=1,
    sm="CONSTANT"
)

print(result)

R Counterpart

• Closest R function: TraMineRextras seqCompare
• Mapping note: Same one-group vs two-group comparison logic.

Notes

• group and seqdata2 cannot both be None.
• Only two groups are supported among valid cases.
• Missing values in group (or in set_var when used) are removed before testing.

Authors

Code: Yuqi Liang

Documentation: Yuqi Liang

References

Liao, T. F., & Fasang, A. E. (2021). Comparing groups of life-course sequences using the Bayesian information criterion and the likelihood-ratio test. Sociological Methodology, 51(1), 44-85.