[FIXED] How to vertically align Vega-Lite stacked bar chart data labels

Issue

How can I align data labels in a stacked bar chart to be centered vertically within each bar segment? In the following example, you can see that the text is positioned at the top of each bar segment. Where a bar segment is thin, the data label overlaps the one below. I realize that having multiple adjacent thin segments would result in overlap even if labels were centered vertically, but that case is unlikely with my data set.

{
  "$schema": "https://vega.github.io/schema/vega-lite/v2.json",
  "data": {
    "values": [
      {"Value": 0.321, "Date": "09/30/2021", "Measure": "Measure 4"},
      {"Value": 0.031, "Date": "09/30/2021", "Measure": "Measure 3"},
      {"Value": 0.123, "Date": "09/30/2021", "Measure": "Measure 2"},
      {"Value": 0.475, "Date": "09/30/2021", "Measure": "Measure 1"}
    ]
  },
  "width": 500,
  "height": 250,
  "resolve": {"scale": {"color": "independent"}},
  "layer": [
    {
      "mark": "bar",
      "encoding": {
        "y": {
          "field": "Value",
          "type": "quantitative",
          "axis": {"format": ".1%"}
        },
        "x": {"field": "Date", "type": "nominal", "axis": {"labelAngle": -45}},
        "color": {"field": "Measure", "type": "nominal"}
      }
    },
    {
      "mark": {"type": "text"},
      "encoding": {
        "y": {"field": "Value", "type": "quantitative", "stack": "zero"},
        "x": {"field": "Date", "type": "nominal"},
        "color": {
          "field": "Measure",
          "type": "nominal",
          "scale": {"range": ["white"]},
          "legend": null
        },
        "text": {
          "aggregate": "sum",
          "field": "Value",
          "type": "quantitative",
          "format": ".1%"
        }
      }
    }
  ]
}

Solution

You can do this using a stack and calculate transform. The text layer would look like this:

    {
      ...
      "transform": [
        {"stack": "Value", "groupby": ["Date"], "as": ["lower", "upper"]},
        {"calculate": "(datum.lower + datum.upper) / 2", "as": "midpoint"}
      ],
      "encoding": {
        "y": {"field": "midpoint", "type": "quantitative"},
        ...
      }
    }

You can see the full spec in the vega editor:

Answered By - jakevdp

Answer Checked By - Dawn Plyler (PHPFixing Volunteer)

Read More

[FIXED] How to control decimal places for value labels in bar charts

Issue

Consider the figure below:

This question is about how to keep all the value labels in the stacked bars in one decimal place, to allow for consistency. Thus, to present -1, 2 and 5 as -1.0, 2.0 and 5.0 in line with others. Here is an example data

df  <- data.frame(group=c("satisfied", "satisfied",
                      "unsatisfied","unsatisfied",
                      "satisfied", "satisfied", 
                      "unsatisfied","unsatisfied"
                      ), 
                cost=c("low","high","low","high",
                      "low","high","low","high"),     
                treatment=c("treated","treated","treated","treated",
                            "untreated","untreated",
                      "untreated","untreated") ,
                value=c(2.3,8.7,5.0,3.1,9.4,3.1,2.0,-1.0)) 

and the code for generating the figure is

#REORDER
df$group <- factor(df$group, levels = c("satisfied", 
                            "unsatisfied"))

ggplot(data=df,aes(y = value, x = group, fill = cost)) +
geom_bar(stat="identity",position='stack') + 
ylab("Y label") +
theme(legend.direction = "horizontal",legend.position = "bottom",
    legend.spacing.x = unit(0.1, 'cm'))+
theme(legend.title=element_blank())+
geom_text(aes(label = ifelse(value !=0, value, "")), 
         position = position_stack(vjust=0.5))+
facet_grid( ~ treatment)

Following how to put exact number of decimal places on label ggplot bar chart I attempted to resolve this by introducing

sprintf("%0.1f", round(value, digits = 2))

into the ggplot function, but this does not produce the desired output. I greatly appreciate any help on this.

Solution

The problem is that you rounded (and thus converted to character, losing numeric information) before using sprintf. To format numbers, you need to give sprintf numbers, not strings. It will take care of the rounding itself. Try this:

label = ifelse(value !=0, sprintf("%0.1f", value), "")

Making the whole code:

ggplot(data = df, aes(y = value, x = group, fill = cost)) +
  geom_bar(stat = "identity", position = 'stack') +
  ylab("Y label") +
  theme(
    legend.direction = "horizontal",
    legend.position = "bottom",
    legend.spacing.x = unit(0.1, 'cm')
  ) +
  theme(legend.title = element_blank()) +
  geom_text(aes(label = ifelse(value !=0, sprintf("%0.1f", value), "")),
            position = position_stack(vjust = 0.5)) +
  facet_grid(~ treatment)

---

The above is a little weird because of the ifelse. A more standard ggplot2 solution would have you get rid of the 0s another way - maybe filter it out before plotting, give data = filter(df, y != 0) to ggplot(). Then you can use the scales functions

label = scales::label_number(accuracy = 0.1)(value)

Making the whole code as below, for the same result:

ggplot(data = dplyr::filter(df, value != 0), aes(y = value, x = group, fill = cost)) +
  geom_bar(stat = "identity", position = 'stack') +
  ylab("Y label") +
  theme(
    legend.direction = "horizontal",
    legend.position = "bottom",
    legend.spacing.x = unit(0.1, 'cm')
  ) +
  theme(legend.title = element_blank()) +
  geom_text(aes(label = scales::label_number(accuracy = 0.1)(value)),
            position = position_stack(vjust = 0.5)) +
  facet_grid(~ treatment)

Answered By - Gregor Thomas

Answer Checked By - Marilyn (PHPFixing Volunteer)

Read More