Visualizing sequences

… ggplotify sequence data plots

Click here to get instructions… 
# assuming you are working within .Rproj environment
library(here)

# install (if necessary) and load other required packages
source(here("source", "load_libraries.R"))

# load environment generated in "2-0_ChapterSetup.R"
load(here("data", "R", "2-0_ChapterSetup.RData"))

 

Virtually all figures shown in the book were rendererd with TraMineR::seqplot which uses base R’s plot function. While there is nothing wrong with that, many R users tend to prefer {ggplot2} for visualizing their data.

We wrote a little R package named {ggseqplot} that reproduces sequence plots from {TraMineR}’s seqplot using {ggplot2}. These plots are produced on the basis of a sequence object defined with TraMineR::seqdef. The package automates the reshaping and plotting of sequence data required to render the plots with {ggplot2}.

{ggseqplot} has been written after this book has been published. You can install and load it with the following commands:

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

For further information we refer to the {ggseqplot} website and a blog post on the website of the Sequence Analysis Association (SAA).

The following section briefly illustrates the package by rendering sequence distribution and index plots.

 

State distribution plot

In the following example we are rendering state distribution plots of the sequence object partner.child.year.seq.

We start by comparing seqdplot and ggseqdplot in their most basic specification:

seqdplot(partner.child.year.seq)

ggseqdplot(partner.child.year.seq)

Note that the default labeling of the x-axis values might produce non-ideal results. The default labeling behavior forces labels to appear at the first and the last position of the sequence data. The states in-between are labeled according to base R’s pretty function.

Of course, the labels can also be adjusted according to your need using the regular {ggplot2} functions. If you want to access the value labels for the x-axis, you can either get them from the sequence object you are plotting (attributes(partner.child.year.seq)$names) or by saving your ggseqdplot as an object and then extracting the required information from the data column k.

In the following example we utilize ggplot2::guide_axis to provide enough space for printing the complete set of values labels for the x-axis. Overlapping labels are avoided by printing the labels in two rows (n.dodge = 2):

dplot <- ggseqdplot(partner.child.year.seq)

dplot + scale_x_discrete(labels = levels(dplot$data$k),
                         guide = guide_axis(n.dodge = 2))

Alternatively, you could get rid of the long value labels altogether and just print the sequence position as labels on the x-axis.

ggseqdplot(partner.child.year.seq) +
  scale_x_discrete()

As ggseqdplot has been written with the intention to mimic the behavior of seqdplot it’s not very surprising that both plots look very similar. Just like with seqplot the ggplot version of the plot allows to turn off weights and to generate faceted plots by specifying the group argument. Turn to the documentation for further possibilities to adjust the outcome of ggseqdplot:

ggseqdplot(partner.child.year.seq, 
           group = as_factor(family$sex), # `as_factor` not necessary; only used to get nice labels
           weighted = FALSE) +
  scale_x_discrete()

Note that it is not possible to obtain the faceted output by manually specifying facet_wrap(~family$sex). That is because ggseqdplot internally reshapes the sequence data into an aggregated format that is suitable for generating the plot. In our example the resulting dataset includes one row for every combination of sequence position \(k\) (22 positions in our case) and state of the alphabet \(A\) (9 distinct states). With one group this yields a data set with \(22*9=198\) rows. The data for the faceted version shown above has 396 rows with 198 for men and women each. The grouping variable family$sex, however, is of length 1866 (one row for each case).

If you want to access the data used for the plot you can save the plot as a list object:

dplot <- ggseqdplot(partner.child.year.seq,
                    group = as_factor(family$sex), 
                    weighted = FALSE)

dplot$data
# A tibble: 396 × 6
   group  state            k        x     value grouplab       
   <fct>  <fct>            <fct>    <fct> <dbl> <glue>         
 1 Female Single, no child state1   1     0.494 Female (n=1039)
 2 Female Single, no child state13  2     0.409 Female (n=1039)
 3 Female Single, no child state25  3     0.330 Female (n=1039)
 4 Female Single, no child state37  4     0.276 Female (n=1039)
 5 Female Single, no child state49  5     0.244 Female (n=1039)
 6 Female Single, no child state61  6     0.218 Female (n=1039)
 7 Female Single, no child state73  7     0.183 Female (n=1039)
 8 Female Single, no child state85  8     0.140 Female (n=1039)
 9 Female Single, no child state97  9     0.133 Female (n=1039)
10 Female Single, no child state109 10    0.122 Female (n=1039)
# … with 386 more rows
# ℹ Use `print(n = ...)` to see more rows

Of course, the figures produced with {ggseqplot} can be adjusted like any other {ggplot2} figure

library(hrbrthemes)

ggseqdplot(partner.child.year.seq) + 
  scale_x_discrete(labels = 18:39) + 
  labs(title = "State distribution plot",
       x = "Age") +
  guides(fill=guide_legend(title="Alphabet")) +
  theme_ipsum_rc() +
  theme(plot.title = element_text(size = 34, 
                                  margin=margin(0,0,20,0)),
        plot.title.position = "plot",
        axis.title.x = element_text(size=12),
        axis.title.y = element_text(size=12))

Sequence index plots

We again start with a comparison of the default outputs for seqIplot and ggseqiplot. We plot a sample of 50 sequences and sort the remaining sequences according to the time spent in marriage (states 7 to 9).

set.seed(1234)
subset <- sample(1:nrow(partner.child.year.seq),50)

seqIplot(partner.child.year.seq[subset,], 
         sortv = rowSums(seqistatd(partner.child.year.seq[subset,])[,7:9]))

ggseqiplot(partner.child.year.seq[subset,],
           sortv = rowSums(seqistatd(partner.child.year.seq[subset,])[,7:9]))

The following code chunk illustrates how the plot can be fine-tuned in the well-known {ggplot2} fashion:

ggseqiplot(partner.child.year.seq[subset,],
           sortv = rowSums(seqistatd(partner.child.year.seq[subset,])[,7:9])) +
  scale_y_continuous(expand = expansion(add = c(.8, 0))) +
  scale_x_discrete(breaks = c(seq(1,21,2),22),
                     labels = c(seq(18,38,2),39)) +
  labs(title = "Sequence index plot",
       x = "Age", 
       fill = "Alphabet",
       caption = "Sequences sorted by marriage duration") +
  guides(color="none") +
  theme_ipsum_rc() +
  theme(plot.title = element_text(size = 34, 
                                  margin=margin(0,0,20,0)),
        plot.title.position = "plot",
        plot.caption = element_text(size=12),
        axis.title.x = element_text(size=12),
        axis.title.y = element_text(size=12))

Corrections

If you see mistakes or want to suggest changes, please create an issue on the source repository.

Reuse

Text and figures are licensed under Creative Commons Attribution CC BY-NC 4.0. Source code is available at https://github.com/sa-book/sa-book.github.io, unless otherwise noted. The figures that have been reused from other sources don't fall under this license and can be recognized by a note in their caption: "Figure from ...".