Contiguous Motif Workflow
In gp3sequences: Transparent Analysis of Ordered Categorical Sequences

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.width = 7,
  fig.height = 4.5
)

Scope

This article demonstrates the restricted contiguous-motif workflow in gp3sequences. The workflow accepts ordinary long-format data frames and does not require Gazepoint software, hardware, exports, or gp3tools.

The functions describe recurring state windows and their locations. They do not infer attention, cognition, emotion, intention, psychological state, or causal mechanisms.

Synthetic ordered-state data

The example contains three sequences and one preserved grouping variable.

library(gp3sequences)

sequence_data <- data.frame(
  sequence = c(
    rep("s1", 5L),
    rep("s2", 4L),
    rep("s3", 3L)
  ),
  order = c(1:5, 1:4, 1:3),
  state = c(
    "A", "B", "A", "B", "C",
    "A", "B", "C", "B",
    "B", "A", "B"
  ),
  group = c(
    rep("g1", 5L),
    rep("g1", 4L),
    rep("g2", 3L)
  ),
  stringsAsFactors = FALSE
)

sequence_data

Extract contiguous motifs

extract_sequence_ngrams() enumerates contiguous state windows only. Minimum and maximum motif lengths and the overlapping-occurrence policy are explicit.

extracted <- extract_sequence_ngrams(
  sequence_data,
  sequence_id_col = "sequence",
  order_col = "order",
  state_col = "state",
  metadata_cols = "group",
  min_length = 2L,
  max_length = 3L,
  overlap = "allow"
)

head(extracted$occurrences)

Summarise, filter, and format motifs

Sequence prevalence uses every validated sequence as its denominator. Filtering is deterministic and retains explicit thresholds and tie handling.

motif_summary <- summarise_sequence_motifs(extracted)

motif_filter <- filter_sequence_motifs(
  motif_summary,
  min_occurrences = 2L,
  min_sequences = 2L,
  min_prevalence = 0.50,
  motif_lengths = c(2L, 3L),
  top_n = 10L,
  rank_by = "sequence_prevalence",
  ties = "include"
)

motif_table <- format_sequence_motifs(
  motif_filter,
  prevalence = "percent",
  digits = 1L
)

motif_table$table

Summarise motif positions

Positions may represent the start, centre, or end of each motif occurrence. Absolute positions use one-based state indices. Relative positions range from 0 to 1 across each sequence.

position_summary <- summarise_sequence_motif_positions(
  extracted,
  position = "centre",
  scale = "relative",
  by = "group"
)

position_summary$summary

format_sequence_motif_positions() changes display precision and units without modifying the underlying analytical object.

position_table <- format_sequence_motif_positions(
  position_summary,
  digits = 1L,
  position_units = "percent",
  include_rank = TRUE
)

position_table$table

Plot motif prevalence

plot_sequence_motifs() uses base R graphics. The returned data frame contains the exact motifs and values used in the plot.

plotted_motifs <- plot_sequence_motifs(
  motif_summary,
  metric = "sequence_prevalence",
  top_n = 8L,
  motif_lengths = c(2L, 3L),
  ties = "include",
  horizontal = TRUE
)

plotted_motifs[
  c("plot_rank", "motif", "motif_length", "plot_value")
]

Plot motif locations

The strip display shows individual occurrence positions with deterministic stacking. No random jitter is used.

strip_data <- plot_sequence_motif_positions(
  extracted,
  position = "centre",
  scale = "relative",
  top_n = 5L,
  display = "strip"
)

head(
  strip_data[
    c(
      "sequence_id",
      "motif",
      "position_value",
      "plot_rank"
    )
  ]
)

The distribution display provides a compact base-R summary for the same structural positions.

plot_sequence_motif_positions(
  extracted,
  position = "centre",
  scale = "relative",
  top_n = 5L,
  display = "distribution"
)

Interpretation boundary

The reported counts, prevalence values, and positions describe the supplied ordered states under the declared preparation, motif-length, overlap, filtering, and position rules. Any substantive interpretation belongs to the research design and cannot be inferred automatically from motif structure alone.