reorder_var: Re-order the columns of a data frame based on MDS or ANOVA
Description
For the MDS method, we use (1 - correlation matrix) as the distance matrix
and re-order the columns according their distances between each other (i.e.
1-dimension representation of p-dimension); as a result, columns in a
neighborhood indicate they are more similar to each other. For the ANOVA
method, if there is a column named .color, it will be used as the x
variable and we perform ANOVA on each numeric column vs this variable, then
the columns are re-ordered by the P-values, so the colors can discriminate
the first few columns most apart. Of course, when there is only a single
color in the .color variable, the ANOVA method will not work and the
original order will be returned. For the randomForest method, the variables
will be ordered by the importance scores (mean descrease in accuracy) and the
argument x will be used as the response variable.
Usage
reorder_var(data, type = c("none", "MDS", "ANOVA", "randomForest"), vars = names(data), numcol = sapply(data, is.numeric), x = data$.color)
Arguments
data
a data frame (or similar data structures like mutaframes)
type
the method to re-order the variables (columns)
vars
the column names of the data
numcol
a logical vector indicating which columns are numeric
x
the x variable to be used in ANOVA and randomForest
Value
the column names (i.e. the argument vars) after being
re-ordered; note non-numeric variables will always be put in the end and
they will not go into the computation
reorder_var(tennis, type = "MDS")
reorder_var(iris, type = "ANOVA", x = iris$Species)
names(iris) # original column namesreorder_var(iris, type = "randomForest", x = iris$Species)