Palmer Penguins¶

This notebook is inspired by an example Radar chart with ggradar.

In [1]:

%useLatestDescriptors
%use dataframe
%use lets-plot

In [2]:

fun rescaleInGroupDataFrame(df: DataFrame<*>, valueCol: String, groupCol: String, rescaledCol: String? = null): DataFrame<*> {
    val rescaledColName = if (rescaledCol == null) {
        "rescaled_${valueCol}"
    } else {
        rescaledCol
    }
    fun rescaleSubDataFrame(subDf: DataFrame<*>): DataFrame<*> {
        val minValue = subDf.minByOrNull(valueCol)?.let { it[valueCol] } as Double
        val maxValue = subDf.maxByOrNull(valueCol)?.let { it[valueCol] } as Double
        return subDf.add(rescaledColName) { (valueCol<Double>() - minValue) / (maxValue - minValue) }
    }
    return df.select(groupCol).distinct().map { v -> rescaleSubDataFrame(df.filter { groupCol<String>() == v[groupCol] }) }.concat()
}

fun getData(): DataFrame<*> {
    val df = DataFrame.readCSV("https://raw.githubusercontent.com/JetBrains/lets-plot-docs/master/data/penguins.csv")
        .dropNulls()
        .rename("bill_length_mm", "bill_depth_mm", "flipper_length_mm", "body_mass_g")
            .into("avg. bill length", "avg. bill depth", "avg. flipper length", "avg. body mass")
        .groupBy("species")
        .mean()
        .gather("avg. bill length", "avg. bill depth", "avg. flipper length", "avg. body mass")
            .into("variable", "value")
        .update { "value"<Double>() }
            .where { "variable"<String>() == "avg. body mass" }
            .with { it / 1000 }
        .add("units") {
            when ("variable"<String>()) {
                "avg. body mass" -> "kg"
                else -> "mm"
            }
        }
    return rescaleInGroupDataFrame(df, "value", "variable")
        .add("rescaled_value_pct") { floor(100 * "rescaled_value"<Double>()) }
        .convert { "rescaled_value_pct"<Double>() }.to<Int>()
        .sortBy("species")
}

In [3]:

val df = getData()
df.head(4)

Out[3]:

DataFrame: rowsCount = 4, columnsCount = 6

species	variable	value	units	rescaled_value	rescaled_value_pct
Adelie	avg. bill length	38.823973	mm	0.000000	0
Adelie	avg. bill depth	18.347260	mm	0.978584	97
Adelie	avg. flipper length	190.102740	mm	0.000000	0
Adelie	avg. body mass	3.706164	kg	0.000000	0

In [4]:

val fontFamily = "roboto"
val axisColor = "lightgray"
val axisTextData = mapOf(
    "x" to List(3) { "avg. bill length" },
    "y" to listOf(0, 50, 100),
    "text" to listOf("0%", "50%", "100%")
)
val penguinsColors = mapOf(
    "Adelie" to "#ff5a5f",
    "Chinstrap" to "#ffb400",
    "Gentoo" to "#007a87"
)
val penguinsTooltips = layerTooltips().title("@species").line("@variable (@units): @value").format("@value", ".2~f")
val penguinsTheme = theme(
    text = elementText(family = fontFamily, size = 18),
    plotTitle = elementText(size = 28, hjust = .5, face = "bold"),
    axisTitle = "blank", axisTextY = "blank", axisLineX = "blank",
    axisTicks = elementLine(color = axisColor),
    panelGrid = elementLine(color = axisColor),
    panelInset = Pair(0, 100),
    tooltip = elementRect(),
    axisTooltip = "blank",
).legendPosition(1, 0).legendJustification(1, 0)

In [5]:

letsPlot(df.toMap()) +
    geomArea(position = positionIdentity, flat = true,
             size = 2.5, colorBy = "paint_a", fillBy = "paint_a", alpha = .2)
        { x = "variable"; y = "rescaled_value_pct"; paint_a = "species" } +
    geomPoint(size = 6, colorBy = "paint_a", tooltips = penguinsTooltips)
        { x = "variable"; y = "rescaled_value_pct"; paint_a = "species" } +
    geomText(data = axisTextData, hjust = 1, fontface = "bold", family = fontFamily, size = 10)
        { x = "x"; y = "y"; label = "text" } +
    scaleXDiscrete() +
    scaleManual("paint_a", name = "", values = penguinsColors) +
    coordPolar(ylim = Pair(-15, 100)) +
    ggsize(800, 600) +
    ggtitle("Penguins species") +
    penguinsTheme + flavorSolarizedLight()

Out[5]: