Reordering Bars in geom_bar ggplot2 by Value

Introduction

In data visualization, the ordering of bar charts is crucial for effective information communication. ggplot2, as one of the most popular plotting packages in R, provides powerful sorting capabilities. This article delves into how to use the reorder() function for custom sorting of bar charts, with particular focus on descending order from high to low.

Problem Analysis

Consider a dataset containing miRNA expression values with the following structure:

miRNA         variable value
mmu-miR-532-3p      pos     7
mmu-miR-1983        pos    75
mmu-miR-301a-3p     pos    70
mmu-miR-96-5p       pos     5
mmu-miR-139-5p      pos    10
mmu-miR-5097        pos    47

Using standard plotting code:

ggplot(corr.m, aes(x = reorder(miRNA, value), y = value, fill = variable)) + 
  geom_bar(stat = "identity")

produces a bar chart ordered by ascending values, which is often not the most effective visualization approach since important high-value items are placed at the end of the chart.

Solution

To achieve descending order from high to low, a negative sign must be added to the value variable within the reorder() function:

ggplot(corr.m, aes(x = reorder(miRNA, -value), y = value, fill = variable)) + 
  geom_bar(stat = "identity")

The principle behind this method is that the reorder() function defaults to rearranging factor levels in ascending order of the specified variable. By adding a negative sign, we effectively sort in descending order because larger positive numbers become smaller negative numbers when negated.

Technical Details

The reorder() function works by rearranging the levels of the first parameter (a factor) based on the second parameter (a numeric vector). In ggplot2, the factor level order of the x-axis variable directly determines the arrangement of bars.

Data processing example:

corr.m <- structure(list(
  miRNA = structure(c(5L, 2L, 3L, 6L, 1L, 4L), 
    .Label = c("mmu-miR-139-5p", "mmu-miR-1983", "mmu-miR-301a-3p", 
               "mmu-miR-5097", "mmu-miR-532-3p", "mmu-miR-96-5p"), 
    class = "factor"),
  variable = structure(c(1L, 1L, 1L, 1L, 1L, 1L), 
    .Label = "pos", class = "factor"),
  value = c(7L, 75L, 70L, 5L, 10L, 47L)
), class = "data.frame", row.names = c("1", "2", "3", "4", "5", "6"))

Extended Applications

Beyond simple descending order, the reorder() function can be combined with other functions to implement more complex sorting logic. For example, custom functions can be used to calculate sorting criteria:

# Sort by squared values
ggplot(corr.m, aes(x = reorder(miRNA, value^2), y = value, fill = variable)) + 
  geom_bar(stat = "identity")

This approach is particularly useful when dealing with nonlinear relationships or data requiring special sorting criteria.

Best Practices

In data visualization projects, it is recommended to always consider bar chart sorting strategies:

• For comparative analysis, prioritize descending order to highlight important items
• Consider using colors and labels to enhance readability
• When dealing with multiple variables, use faceting to organize complex sorting requirements

Conclusion

By understanding how the reorder() function works and the sorting mechanisms in ggplot2, data scientists can create more effective and intuitive visualizations. The negative sign technique is a simple yet powerful tool that quickly achieves descending order, thereby improving data communication effectiveness.