In addition to imputation methods, VIM provides a number
of functions, which can be used to plot results in sophisticated
ways.
This vignette showcases selected plotting functions, which are very supportive in context with visualizing missing and imputed values.
The following example demonstrates the functionality of the plotting
functions using a subset of sleep. In order to emphasize
the features of the methods, the missing values in the dataset are
imputed via kNN() or regressionImp(). Both are
powerful donor-based imputation methods and also included in the
VIM package. (see vignette("donorImp"))
library(VIM)
library(magrittr)
dataset <- sleep[, c("Dream", "NonD", "BodyWgt", "Span")] # dataset with missings
dataset$BodyWgt <- log(dataset$BodyWgt)
dataset$Span <- log(dataset$Span)
imp_knn <- kNN(dataset) # dataset with imputed valuesTo keep things as simple as possible, the plotting functions in
VIM uses three main colors. Each color represents a
property:
The aggr() function calculates or plots the amount of
missing/imputed values in each variable and the amount of
missing/imputed values in certain combinations of variables.
aggr(dataset) 
aggr(imp_knn, delimiter = "_imp") 
The plots indicate that all missing values in the dataset are imputed
via knn(). (All the previously red bars changed their color
to orange)
The barMiss() function provides a barplot with
highlighting of missing/imputed values in other variables by splitting
each bar into two parts. Additionally, information about missing/imputed
values in the variable of interest is shown on the right hand side.
If only.miss=TRUE, the missing/imputed values in the
variable of interest are visualized by one bar on the right hand side.
If additional variables are supplied, this bar is again split into two
parts according to missingness/number of imputed missings in the
additional variables.
# for missing values
x <- sleep[, c("Exp", "NonD", "Sleep")]
barMiss(x, only.miss = FALSE)
#>
#> Click in in the left margin to switch to the previous variable or in the right margin to switch to the next variable.
#> To regain use of the VIM GUI and the R console, click anywhere else in the graphics window.
# for imputed values
x_IMPUTED <- regressionImp(NonD ~ Sleep, data = x)
#> There still missing values in variable NonD . Probably due to missing values in the regressors.
barMiss(x_IMPUTED, delimiter = "_imp", only.miss = FALSE)
#>
#> Click in in the left margin to switch to the previous variable or in the right margin to switch to the next variable.
#> To regain use of the VIM GUI and the R console, click anywhere else in the graphics window.The plot indicates that there are still some missings in NonD. This is because the regression model could not be applied to observations, where Sleep is unobserved.
In addition to a standard scatterplot, lines are plotted in
scattMiss() for the missing values in one variable. If
there are imputed values, they will be highlighted.
Information about missing values in one variable is included as
vertical or horizontal lines, as determined by the side
argument. The lines are thereby drawn at the observed x- or y-value. In
case of imputed values, they will additionally be highlighted in the
scatterplot. Supplementary, percentage coverage ellipses can be drawn to
give a clue about the shape of the bivariate data distribution.
In contrast to the other examples, regressionImp() is
used for imputing missing values. This has been done deliberately to
highlight the functionality of scattMiss(). The following
plots makes it easy to indentify missing/imputed values.
dataset <- sleep[, c("Span", "NonD","Sleep")]
# for missing values
scattMiss(dataset[,-3])
#>
#> Click in bottom or left margin to change the 'side' argument accordingly.
#> To regain use of the VIM GUI and the R console, click anywhere else in the graphics window.
# for imputed values
imp_regression <- regressionImp(NonD ~ Sleep, dataset)
#> There still missing values in variable NonD . Probably due to missing values in the regressors.
scattMiss(imp_regression[,-3], delimiter = "_imp")
#>
#> Click in bottom or left margin to change the 'side' argument accordingly.
#> To regain use of the VIM GUI and the R console, click anywhere else in the graphics window.The plot indicates that there are still some missings in
NonD. This is because the regression model could not be
applied to observations, where Sleep is unobserved.
The histMiss() function visualizes data in a histogram
with highlighting the missing/imputed values in other variables by
splitting each bin into two parts. Additionally, information about
missing/imputed values in the variable of interest is shown on the right
hand side.
If only.miss=TRUE, the missing/imputed values in the
variable of interest are visualized by one bar on the right hand side.
If additional variables are supplied, this bar is again split into two
parts according to missingness/number of imputed missings in the
additional variables.
## for missing values
x <- sleep[, c("Span", "NonD","Sleep")]
histMiss(x, only.miss = FALSE)
#>
#> Click in in the left margin to switch to the previous variable or in the right margin to switch to the next variable.
#> To regain use of the VIM GUI and the R console, click anywhere else in the graphics window.
# for imputed values
x_IMPUTED <- regressionImp(NonD ~ Sleep, data = x)
#> There still missing values in variable NonD . Probably due to missing values in the regressors.
histMiss(x_IMPUTED, delimiter = "_imp", only.miss = FALSE)
#>
#> Click in in the left margin to switch to the previous variable or in the right margin to switch to the next variable.
#> To regain use of the VIM GUI and the R console, click anywhere else in the graphics window.The matrixplot() function creats a matrix plot, in which
all cells of a data matrix are visualized by rectangles. Available data
is coded according to a continuous color scheme, while missing/imputed
data is visualized by a clearly distinguishable color.
x <- sleep[, c("Dream", "NonD","Sleep", "BodyWgt")]
x$BodyWgt <- log(x$BodyWgt)
# for missing values
matrixplot(x, sortby="BodyWgt")
#>
#> Click in a column to sort by the corresponding variable.
#> To regain use of the VIM GUI and the R console, click outside the plot region.
# for imputed values - multiple variable imputation with regrssionImp()
x_IMPUTED <- regressionImp(NonD + Dream ~ Sleep, data = x)
#> There still missing values in variable NonD . Probably due to missing values in the regressors.
#> There still missing values in variable Dream . Probably due to missing values in the regressors.
matrixplot(x_IMPUTED, delimiter = "_imp", sortby = "BodyWgt")
#>
#> Click in a column to sort by the corresponding variable.
#> To regain use of the VIM GUI and the R console, click outside the plot region.In addition to a standard scatterplot, information about missing/imputed values is shown in the plot margins. Furthermore, imputed values are highlighted in the scatterplot.
Boxplots for available and missing/imputed data, as well as univariate scatterplots for missing/imputed values in one variable are shown in the plot margins.Imputed values in either of the variables are highlighted in the scatterplot.
Furthermore, the frequencies of the missing/imputed values can be displayed by a number (lower left of the plot). The number in the lower left corner is the number of observations that are missing/imputed in both variables.
dataset <- sleep[, c("Dream", "NonD", "BodyWgt", "Span")]
dataset$BodyWgt <- log(dataset$BodyWgt)
dataset$Span <- log(dataset$Span)
imp_knn <- kNN(dataset, variable = "NonD")
dataset[, c("NonD", "Span")] %>%
marginplot()
imp_knn[, c("NonD", "Span", "NonD_imp")] %>%
marginplot(delimiter = "_imp")
The marginmatrix() function creates a scatterplot matrix
with information about missing/imputed values in the plot margins of
each panel.
## for missing values
x <- sleep[, 2:4]
x[, 1] <- log10(x[, 1])
marginmatrix(x)
## for imputed values
x_imp <- irmi(sleep[, 2:4])
x_imp[,1] <- log10(x_imp[, 1])
marginmatrix(x_imp, delimiter = "_imp")