{ggseqplot}

ggplotify sequence plots (updated slides)

Motivation

{TraMineR}’s default plots appearance almost publication-ready
But: virtually always some adjustments are necessary
Requires some knowledge of base R’s {plot}
However, today most (new) R users prefer {ggplot2}

Motivation

Goals:
- Provide a package that uses {ggplot2} to render sequence plots
- and that allows for using {ggplot2} functions and extensions to change the plot appearance
Tasks:
- reshape sequence data into format required by {ggplot2} (usually long data format)
- use appropriate geom_* functions to rebuild seqplot functions

install.packages("ggseqplot")
library(ggseqplot)

What we can do with {ggseqplot}

Summarization plots	TraMineR functions	ggseqplot function	ggplot2 function
State Distribution Plot	`seqdplot` `seqstatd`	`ggseqdplot`	`geom_bar`
Entropy Line Plot	`seqHtplot` `seqstatd`	`ggseqeplot`	`geom_line`
Modal State Sequence Plot	`seqmsplot` `seqmodst`	`ggseqmsplot`	`geom_bar`
Mean Time Plot	`seqmtplot` `seqmeant`	`ggseqmtplot`	`geom_bar`
Transition Rate Plot	`seqtrate`	`ggseqtrplot`	`geom_tile`

What we can do with {ggseqplot}

Representation plots	TraMineR functions	ggseqplot functions	ggplot2 and related functions
Sequence Index Plot	`seqiplot`	`ggseqiplot`	`geom_rect` `ggh4x::facetted_pos_scales`
Sequence Frequency Plot	`seqfplot`	`ggseqfplot` `(ggseqiplot)`	`seqtab` `geom_rect` `ggh4x::facetted_pos_scales`
Representative Sequence Plot	`seqrplot` `seqrep`	`ggseqrplot` `(ggseqiplot)`	`geom_rect` `ggrepel::geom_label_repel` `ggrepel::geom_text_repel` `ggtext::element_markdown` `patchwork`
Relative Frequency Sequence Plot	`seqrfplot`	`ggseqrfplot` `(ggseqiplot)`	`geom_rect` `geom_boxplot` `patchwork`

Example

We use the well-known example data from {TraMineR} to render some plots.

data(actcal)

set.seed(1)
actcal <- actcal[sample(nrow(actcal),300),]

actcal.lab <- c(
  "> 37 hours", "19-36 hours", 
  "1-18 hours", "no work"
  )

actcal.seq <- seqdef(
  actcal,
  13:24,
  labels = actcal.lab
  )

State distribution plots

seqdplot(actcal.seq)

ggseqdplot(actcal.seq)

State distribution plots

ggseqdplot(actcal.seq,
           border = TRUE)

State distribution plots

ggseqdplot(actcal.seq,
           border = TRUE) +
  # Built-in months abbreviations for axis labels
  scale_x_discrete(labels = month.abb)

State distribution plots

ggseqdplot(actcal.seq,
           border = TRUE) +
  # Built-in months abbreviations for axis labels
  scale_x_discrete(labels = month.abb) + 
  # change the color palette (fill and border color)
  scale_fill_discrete_sequential("heat")

State distribution plots

ggseqdplot(actcal.seq,
           border = TRUE) +
  # Built-in months abbreviations for axis labels
  scale_x_discrete(labels = month.abb) + 
  # change the color palette (fill and border color)
  scale_fill_discrete_sequential("heat") +
  # apply & adjust alternative theme
  theme_ipsum() +
  theme(
    legend.position = "bottom",
    legend.title = element_blank(),
    legend.text = element_text(size = 11)
    )

State distribution plots

ggseqdplot(actcal.seq, border = TRUE,
           group = actcal$sex,   # Group by gender
           dissect = "row") +    # separate plot for each state
  scale_x_discrete(labels = month.abb) +
  scale_fill_discrete_sequential("heat")

Sequence index plots

seqIplot(actcal.seq, sortv = "from.end")

ggseqiplot(actcal.seq, sortv = "from.end")

Sequence index plots

ggseqiplot(actcal.seq, 
           sortv = "from.end")

Sequence index plots

ggseqiplot(actcal.seq, 
           sortv = "from.end", 
           group = actcal$sex)

Sequence index plots

ggseqiplot(actcal.seq, 
           sortv = "from.end", 
           group = actcal$sex,
           facet_scale = "fixed")

Sequence index plots

# using {ggh4x} to get varying plot sizes

# a vector storing the heights of the subplots   
hghts <- table(fct_drop(actcal$sex))/nrow(actcal.seq)

ggseqiplot(actcal.seq, 
           sortv = "from.end", 
           group = actcal$sex,
           facet_ncol = 1) +
  force_panelsizes(rows = hghts) +
  theme(panel.spacing = unit(1, "lines"))

Sequence index plots

ggseqiplot(actcal.seq, sortv = "from.end") + 
  # Use months abbreviations for axis labels
  scale_x_discrete(labels = month.abb) + 
  # change the fill and border color
  scale_fill_discrete_sequential("heat") +
  scale_color_discrete_sequential("heat") +
  # add a title and a axis title
  labs(x = "Month",
       title = "Piccarreta-flavored Index Plot") +
  # let the time run "bottom-up" instead of "left-right"
  coord_flip() +
  # Change the position and size 
  # of the title and the legend position
  theme(legend.position = "top",
        plot.title = element_text(size = 30),
        plot.title.position = "plot")

Representative sequence index plots

# Compute dissimilarity matrix
lcs.dis <- seqdist(actcal.seq, method="LCS")

seqrplot(actcal.seq, diss = lcs.dis, coverage = .7)

# Compute dissimilarity matrix
lcs.dis <- seqdist(actcal.seq, method="LCS")

ggseqrplot(actcal.seq, diss = lcs.dis, coverage = .7, border = TRUE)

Transition rate plots

ggseqtrplot(actcal.seq, 
            group = actcal$sex)

Some words of caution

Warning

I am working on this package alone in my spare time and I am a novice R developer
Double check ggseqplots by comparing them to the TraMineR plots

Important

Don’t use ggseqrfplot until the current issue is fixed
Dependencies are not specified correctly: update tidyverse package(s)

Some words of caution

Warning

I am working on this package alone in my spare time and I am a novice R developer
Double check ggseqplots by comparing them to the TraMineR plots

Important

~~Don’t use ggseqrfplot until the current issue is fixed~~
~~Dependencies are not specified correctly: update tidyverse package(s)~~
Both issues should be fixed after update to version 0.8.3 (on CRAN since 2023-09-22)

Future plans

High priority: ~~Fix known issues (rfplot; dependencies)~~
Low priority: include additional plot types and revise current functions

maraab23.github.io/ggseqplot

marcel.raab@ifb.uni-bamberg.de