VIM introduces tools for visualization of missing and imputed values.
Forthermore, methods to impute missing values are featured. This
vignette will give a brief look at a common imputation scenario and
showcase how VIM can be used to both impute the data and also interpret
the results visually.
Visualize missing values
library(VIM)
data(sleep)
a <- aggr(sleep, plot = FALSE)
plot(a, numbers = TRUE, prop = FALSE)
The left plot shows the amount of missings for each column in the
dataset sleep and the right plot shows how often each
combination of missings occur. For example, there are 9 rows wich
contain a missing in both NonD and Dream.
For simplicity, we will only look at the variables Dream
and Sleep for the remainer of this vignette. Bivariate
datasets can be passed to special functions that visualize the structure
of missings such as marginplot().
x <- sleep[, c("Dream", "Sleep")]
marginplot(x)
The red boxplot on the left
shows the distrubution of all values of Sleep where
Dream contains a missing value. The
blue boxplot on the left
shows the distribution of the values of Sleep where
Dream is observed.
Impute missing values
In order to impute missing values, VIM offers a spectrum
of imputation methods like kNN() (k nearest neighbour),
hotdeck() and so forth. Those functions can be applied to a
data.frame and return another data.frame where
missings are replaced by imputed values.
To learn more about all implemented imputation methods, three
vignettes are available
vignette("donorImp") explains the donor-based
imputation methods hotdeck() and kNN()vignette("modelImp") gives insight into the model-based
imputation methods regressionImp() and
matchImpute()vignette("irmi") showcases the irmi()
method.
Visualize imputed values
The same functions that visualize missing values can also visualize
the imputed dataset.
marginplot(x_imputed, delimiter = "_imp")
In this plot three differnt colors are used in the top-right. These
colors represent the structure of missings.
- brown points represent
values where
Dream was missing initially
- beige points represent
values where
Sleep was missing initially
- black points represent values where both
Dream and Sleep were missing initially
The kNN() method seemingly preserves the correlation
between Dream and Sleep.