Data visualization

for incomplete datasets in R

Hanne Oberman

Utrecht University

Take aways

Missing data are

  • a pervasive problem
  • visualizable & analyzable
  • informative!

Missingness

A problem to fix

  • unit non-response

\(\rightarrow\) weighting etc.

  • item non-response

\(\rightarrow\) imputation etc.

Case study

set.seed(123)
library(ricu)
library(mice)
library(ggmice)
library(ggplot2)
dat <- mimic_demo |> clean_mimic_demo()

Incomplete data

str(dat)
'data.frame':   50 obs. of  9 variables:
 $ fio2     : num  100 100 100 100 40 50 100 50 100 100 ...
 $ pao2     : num  NA 20 NA NA NA NA NA NA NA NA ...
 $ plt      : num  80 16 189 189 175 127 11 156 87 80 ...
 $ bili     : num  0.9 2.3 0.3 0.3 0.8 NA 1.2 0.6 1.1 2.1 ...
 $ tgcs     : Factor w/ 8 levels "3","4","5","6",..: 1 1 NA NA NA NA NA NA NA NA ...
 $ map      : num  101 94 146 105 94 ...
 $ crea     : num  3.7 1.8 0.5 1.2 2 1.2 1.6 1.6 1 0.9 ...
 $ urine24  : num  296.6 38.1 721.9 1257.6 1516.8 ...
 $ mortality: logi  TRUE TRUE TRUE FALSE FALSE TRUE ...

Incomplete data

visdat::vis_dat(dat)

Response indicator

is.na(dat)
    fio2  pao2   plt  bili  tgcs   map  crea urine24 mortality
1  FALSE  TRUE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
2  FALSE FALSE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
3  FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
4  FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
5  FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
6  FALSE  TRUE FALSE  TRUE  TRUE FALSE FALSE   FALSE     FALSE
7  FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
8  FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
9  FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
10 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE    TRUE     FALSE
11 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
12 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
13 FALSE  TRUE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
14 FALSE FALSE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
15 FALSE  TRUE FALSE  TRUE  TRUE FALSE FALSE   FALSE     FALSE
16 FALSE FALSE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
17 FALSE FALSE FALSE FALSE FALSE FALSE FALSE    TRUE     FALSE
18 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
19 FALSE FALSE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
20 FALSE  TRUE FALSE  TRUE  TRUE FALSE FALSE   FALSE     FALSE
21 FALSE  TRUE FALSE  TRUE FALSE FALSE FALSE   FALSE     FALSE
22 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
23 FALSE  TRUE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
24 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
25 FALSE  TRUE FALSE  TRUE  TRUE FALSE FALSE   FALSE     FALSE
26 FALSE  TRUE FALSE  TRUE  TRUE FALSE FALSE   FALSE     FALSE
27 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
28 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
29 FALSE  TRUE FALSE  TRUE FALSE FALSE FALSE   FALSE     FALSE
30 FALSE  TRUE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
31 FALSE FALSE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
32 FALSE FALSE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
33 FALSE FALSE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
34 FALSE  TRUE FALSE  TRUE  TRUE FALSE FALSE   FALSE     FALSE
35 FALSE FALSE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
36 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
37 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
38 FALSE  TRUE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
39 FALSE FALSE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
40 FALSE  TRUE FALSE  TRUE FALSE FALSE FALSE   FALSE     FALSE
41 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE    TRUE     FALSE
42 FALSE  TRUE FALSE  TRUE FALSE FALSE FALSE   FALSE     FALSE
43 FALSE  TRUE FALSE  TRUE  TRUE FALSE FALSE   FALSE     FALSE
44 FALSE FALSE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
45 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE
46 FALSE  TRUE FALSE  TRUE  TRUE FALSE FALSE   FALSE     FALSE
47 FALSE  TRUE FALSE  TRUE FALSE FALSE FALSE   FALSE     FALSE
48 FALSE FALSE FALSE FALSE FALSE FALSE FALSE   FALSE     FALSE
49 FALSE  TRUE FALSE FALSE  TRUE  TRUE FALSE    TRUE     FALSE
50 FALSE  TRUE FALSE FALSE  TRUE FALSE FALSE   FALSE     FALSE

Response indicator

naniar::vis_miss(dat)

Missingness rate

colSums(is.na(dat))
     fio2      pao2       plt      bili      tgcs       map      crea   urine24 
        0        38         0        13        28         1         0         4 
mortality 
        0

Missingness rate

VIM::aggr(dat, numbers = TRUE, prop = FALSE)

Missing data pattern

mice::md.pattern(dat)
   fio2 plt crea mortality map urine24 bili tgcs pao2   
11    1   1    1         1   1       1    1    1    1  0
5     1   1    1         1   1       1    1    1    0  1
17    1   1    1         1   1       1    1    0    0  2
5     1   1    1         1   1       1    0    1    0  2
8     1   1    1         1   1       1    0    0    0  3
1     1   1    1         1   1       0    1    1    1  1
2     1   1    1         1   1       0    1    0    0  3
1     1   1    1         1   0       0    1    0    0  4
      0   0    0         0   1       4   13   28   38 84

Missing data pattern

ggmice::plot_pattern(dat, square = FALSE)

Coma symptoms

Code 
ggmice(dat, aes(x = tgcs)) +
  geom_bar(fill = "white")

Coma symptoms by mortality

Code 
ggmice(dat, aes(x = tgcs)) +
  geom_bar(fill = "white")  +
  facet_wrap(~factor(
    mortality, 
    levels = c(TRUE, FALSE), 
    labels = c("patient deceased", "patient alive")
    ), ncol = 1)

Blood pressure by mortality

Code 
ggmice(dat, aes(x = map, y = mortality)) +
  geom_point()

Blood oxygenation by mortality

Code 
ggmice(dat, aes(x = pao2, y = mortality)) +
  geom_jitter(width = 0, height = 0.1)

Scatter plot

Code 
ggmice(dat, aes(x = pao2, y = map)) +
  geom_jitter(height = 0)

Faceted scatter plot

Code 
ggmice(dat, aes(x = pao2, y = map)) +
  geom_jitter(height = 0) +
    facet_wrap(~factor(
    mortality, 
    levels = c(TRUE, FALSE), 
    labels = c("patient deceased", "patient alive")
    ), ncol = 1)

Imputation workflow

Imputation models

pred <- make.predictorMatrix(dat)
meth <- make.method(dat)
plot_pred(pred, method = meth, square = FALSE)

Correlation

plot_corr(dat, square = FALSE, label = TRUE)

Imputation models

pred <- quickpred(dat, mincor = 0.3)
plot_pred(pred, method = meth, square = FALSE)

Scatter plot

ggmice(dat, aes(pao2, map)) +
  geom_point(size = 2)

Faceted distribution

ggmice(dat, aes(map)) +
  geom_histogram(fill = "white") +
  facet_grid(factor(
    is.na(pao2), 
    levels = c(TRUE, FALSE), 
    labels = c("missing PaO2", "observed PaO2")
    ) ~ .)

Faceted scatter plot

ggmice(dat, aes(pao2, map)) +
  geom_point(size = 2) +
  facet_wrap(~factor(
    mortality, 
    levels = c(TRUE, FALSE), 
    labels = c("patient deceased", "patient alive")
    ), ncol = 1)

Adjust imputation models

pred["map", c("pao2", "mortality")] <- 1
plot_pred(pred, method = meth, square = FALSE)

Impute

imp <- mice(
  dat, 
  pred = pred, 
  method = meth,
  m = 3,
  seed = 11,
  print = FALSE)
plot_trace(imp, legend = FALSE)

Stripplot of blood pressure

ggmice(imp, aes(x = .imp, y = map)) +
  geom_jitter(width = 0.05) +
  labs(x = "Imputation number")

Stripplot of oxygenation

ggmice(imp, aes(x = .imp, y = pao2)) +
  geom_jitter(width = 0.05) +
  labs(x = "Imputation number")

Blood pressure by mortality

Blood pressure by mortality

Oxygenation by mortality

Oxygenation by mortality

Scatter plot

Scatter plot

Faceted scatter plot

ggmice(dat, aes(pao2, map)) +
  geom_point(size = 2) +
  facet_wrap(~factor(
    mortality, 
    levels = c(TRUE, FALSE), 
    labels = c("patient deceased", "patient alive")
    ), ncol = 1)

Faceted scatter plot

ggmice(imp, aes(pao2, map)) +
  geom_point(size = 2) +
  facet_wrap(~factor(
    mortality, 
    levels = c(TRUE, FALSE), 
    labels = c("patient deceased", "patient alive")
    ), ncol = 1)

Faceted distribution

ggmice(imp, aes(pao2, group = .imp, linetype = !(.imp == 0))) +
  geom_density() +
  facet_wrap(~factor(
    mortality, 
    levels = c(TRUE, FALSE), 
    labels = c("patient deceased", "patient alive")
    ), ncol = 1) +
  scale_linetype(guide = "none")

Take aways

Missing data are

  • a pervasive problem
  • visualizable & analyzable
  • informative!

Thank you!