ManyBabies 1

Meta-analysis

Plot

Forest plot of the meta-analytic results for infant-directed-speech preference. The standardized effect size [x axis] is shown for each lab and age group [y axis]; error bars indicate 95% confidence intervals. Effects are grouped by method [colors]: central fixation, eye tracking, or head-turn preference procedure (HPP). Larger plotted points correspond to greater inverse variance. For each method, the diamond and associated error bars represent the meta-analytic estimate from the method-moderated model and the estimate’s 95% confidence interval. The top diamond shows the global meta-analytic estimate and 95% confidence interval from the unmoderated model.

Plot = import("https://esm.sh/@observablehq/plot@0.6.13")
fontFamily = "Source Sans Pro"
fontSize = ".75rem"

d = transpose(forest_data)

maxX = d3.max(forest_data.effects_cih)
pad = 0.1

methodIndeces = d3.rollup(
  d,
  (values) => ({Method: values[0].Method, order: d3.min(values, (d) => d.order)}),
  (d) => d.Method
)

method_vals = Array.from(methodIndeces.keys()).filter(x => x)

age_sort = (a, b) => parseInt(a.split('-')[0], 10) - parseInt(b.split('-')[0], 10)
age_vals = Array.from(new Set(forest_data.age_group)).filter(d => d).sort(age_sort)

viewof methods = Inputs.checkbox(method_vals, {value: method_vals, label: "Methods"})
viewof ages = Inputs.checkbox(age_vals, {value: age_vals, label: "Age groups"})

methodLabels = Array.from(methodIndeces.values()).filter(x => methods.includes(x.Method))

fd = d.filter(d => methods.includes(d.Method))
      .filter(d => !d.age_group | ages.includes(d.age_group))

Plot.plot({
  style: { fontFamily: fontFamily },
  marginLeft: 155,
  width: 800,
  x: {
    label: "Effect size",
    line: true,
    domain: [d3.min(forest_data.effects_cil) - pad, d3.max(forest_data.effects_cih) + pad],
  },
  y: {
    grid: true,
    line: true,
    label: "Lab and age group",
    labelAnchor: "top",
    padding: 1,
    type: "point",
    fontVariant: (d) => d.meta ? "small-caps" : "normal",
    tickFormat: (value) => fd.find((d) => d.order === value).y_label,
  },
  marks: [
    Plot.ruleX([0], {stroke: "lightgrey", strokeDasharray: [3, 3]}),
    Plot.ruleY(fd.filter(d => d.meta), {
      y: "order",
      stroke: "lightgrey"
    }),
    Plot.ruleX(fd, Plot.pointerY({x: "effects", py: "order", stroke: "Method"})),
    Plot.ruleY(fd, Plot.pointerY({px: "effects", y: "order", stroke: "lightgrey"})),
    Plot.link(fd, {
      y: "order",
      x1: "effects_cil",
      x2: "effects_cih"
    }),
    Plot.dot(fd.filter(d => d.meta), {
      x: "effects",
      y: "order",
      stroke: "Method",
      fill: "white",
      symbol: "diamond",
      strokeWidth: 2,
      r: "inverse_vars",
    }),
    Plot.dot(fd.filter(d => !d.meta), {
      x: "effects",
      y: "order",
      stroke: "Method",
      fill: "Method",
      symbol: "circle",
      strokeWidth: 2,
      r: "inverse_vars",
    }),
    Plot.text(methodLabels.filter((d) => d.Method == "(Global)"), {
      x: maxX + 0.1,
      y: "order",
      text: "Method",
      fill: "Method",
      stroke: "white",
      strokeWidth: 12,
      fontWeight: "bold",
      dy: -6,
      fontSize: 18,
      textAnchor: "end"
    }),
    Plot.text(methodLabels.filter((d) => d.Method !== "(Global)"), {
      x: maxX + 0.1,
      y: "order",
      text: "Method",
      fill: "Method",
      stroke: "white",
      strokeWidth: 12,
      fontWeight: "bold",
      dy: 6,
      fontSize: 18,
      textAnchor: "end"
    }),
    Plot.tip(fd, Plot.pointerY({
      x: "effects",
      y: "order",
      stroke: "Method",
      format: { y: null },
      channels: {
        Lab: "lab",
        "Age group": "age_group",
        "Effect size": (d) => d.effects.toFixed(2),
        CI: (d) => `[${d.effects_cil.toFixed(2)}, ${d.effects_cih.toFixed(2)}]`,
      },
    })),
  ]
})

Data

Aggregated

Plot

The lines on this plot shows fitted values from a mixed effects model predicting looking time [y axis] from trial number [x axis], age [columns], method [rows], trial type [IDS red vs. ADS blue], and language group [NAE dark vs. non-NAE light].

lt = transpose(lt_fits)
age_groups = ["3-6 mo", "6-9 mo", "9-12 mo", "12-15 mo"]
nae_vals = new Map([["North American English (NAE)", "NAE"],
                    ["Not North American English (non-NAE)", "non-NAE"]])
                    
td = transpose(trial_data)

viewof points = Inputs.toggle({label: "Show data points"})

viewof nae = Inputs.checkbox(nae_vals, {value: Array.from(nae_vals.values())})

ltf = lt.filter(d => nae.includes(d.nae))
tdf = td.filter(d => nae.includes(d.nae))

Plot.plot({
  style: { fontFamily: fontFamily },
  marginRight: 90,
  width: 800,
  x: { label: "Trial number", ticks: 4 },
  y: {
    label: "Looking time (seconds)",
    domain: [d3.min(trial_data.looking_time), d3.max(trial_data.looking_time)]
  },
  fx: { label: "", domain: age_groups },
  fy: { label: "" },
  color: {
    legend: true,
    range: ["#CA0020", "#F4A582", "#0571B0", "#92C5DE"], // RdBu for 4 values
    domain: ["IDS, NAE", "IDS, non-NAE", "ADS, NAE", "ADS, non-NAE"]
  },
  marks: [
    Plot.dot(points ? tdf : [], {
      x: "trial_num",
      y: "looking_time",
      fill: "type_nae",
      fx: "age_group",
      fy: "method",
      opacity: 0.4,
      r: 1
    }),
    Plot.line(ltf, {
      x: "trial_num",
      y: "fitted",
      stroke: "type_nae",
      z: (d) => `${d.trial_type} + ${d.nae}`,
      fx: "age_group",
      fy: "method",
    }),
    Plot.tip(ltf, Plot.pointer({
      x: "trial_num",
      y: "fitted",
      stroke: "type_nae",
      z: (d) => `${d.trial_type} + ${d.nae}`,
      fx: "age_group",
      fy: "method",
      format: { x: null, y: null, fx: null, fy: null, stroke: null },
      channels: { "Trial type": "trial_type", "Language group": "nae" },
    })),
  ]
})

Data