juryrig/docs/index.html at main · smkwray/juryrig · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="UTF-8">
  <meta name="viewport" content="width=device-width, initial-scale=1.0">
  <meta name="color-scheme" content="light dark">
  <title>juryrig &mdash; New York Criminal Court Analysis</title>
  <script src="js/theme.js"></script>
  <link rel="stylesheet" href="css/style.css">
  <script src="https://cdn.jsdelivr.net/npm/chart.js@4.4.7/dist/chart.umd.min.js"></script>
</head>
<body>
  <!-- Header -->
  <header class="site-header">
    <div class="header-inner">
      <a href="index.html" class="site-logo">jury<span class="logo-accent">rig</span></a>
      <div class="header-controls">
        <button class="nav-toggle" onclick="document.querySelector('.nav-links').classList.toggle('open')" aria-label="Toggle navigation">&#9776;</button>
        <button id="themeToggle" class="theme-toggle" onclick="juryrigToggleTheme()" aria-label="Toggle theme">&#9790;</button>
      </div>
      <nav>
        <ul class="nav-links">
          <li><a href="index.html" class="active">Home</a></li>
          <li><a href="findings.html">Findings</a></li>
          <li><a href="race-analysis.html">Race Analysis</a></li>
          <li><a href="pretrial.html">Pretrial Impact</a></li>
          <li><a href="methodology.html">Methodology</a></li>
          <li><a href="faq.html">FAQ</a></li>
        </ul>
      </nav>
    </div>
  </header>

  <!-- Hero -->
  <section class="hero">
    <div class="hero-inner">
      <h1>New York Criminal Court Analysis</h1>
      <p class="subtitle">
        Two independent analyses of public New York State court records: a machine-learning model
        estimating conviction likelihood from 1.6 million arraignment records, and a before-after
        comparison of pretrial release rates around bail-law amendments using 854,000 cases.
      </p>
      <div class="stats-row">
        <div class="stat-card">
          <div class="stat-value">1.6M</div>
          <div class="stat-label">Arraignment Records</div>
        </div>
        <div class="stat-card">
          <div class="stat-value">0.86</div>
          <div class="stat-label">Best Model AUROC</div>
        </div>
        <div class="stat-card">
          <div class="stat-value">854K</div>
          <div class="stat-label">Pretrial Cases</div>
        </div>
        <div class="stat-card">
          <div class="stat-value">61</div>
          <div class="stat-label">Counties Analyzed</div>
        </div>
      </div>
    </div>
  </section>

  <!-- Content -->
  <main class="content">

    <!-- Two branches -->
    <section class="section">
      <div class="section-header">
        <h2>Two Research Branches</h2>
      </div>

      <div class="card-grid">
        <div class="card">
          <h3>Branch 1: Conviction Prediction</h3>
          <p>
            Using arraignment-time information alone &mdash; county, charge severity, arrest type,
            case characteristics &mdash; how well can a model estimate whether a case ends in conviction?
          </p>
          <div class="insight-box" style="margin-top: 1rem;">
            <p><strong>Key finding:</strong> Geography dominates. NYC courts have a 25% conviction rate;
            non-NYC courts have 69%. The best model reaches 78% accuracy and AUROC 0.86.</p>
          </div>
          <a href="findings.html" class="card-link">Explore findings</a>
        </div>
        <div class="card">
          <h3>Branch 2: Pretrial Release Impact</h3>
          <p>
            Did pretrial release rates change differently for charge categories targeted by
            New York's May 2022 and June 2023 bail-law amendments?
          </p>
          <div class="insight-box" style="margin-top: 1rem;">
            <p><strong>Key finding:</strong> Firearm charges showed the largest shift &mdash;
            release rates dropped 20&ndash;32 pp more than baseline after May 2022.</p>
          </div>
          <a href="pretrial.html" class="card-link">Explore pretrial impact</a>
        </div>
      </div>
    </section>

    <!-- Quick visual -->
    <section class="section">
      <div class="section-header">
        <h2>At a Glance</h2>
        <span class="section-tag">Interactive</span>
      </div>

      <div class="chart-container">
        <h3>Model Performance Comparison</h3>
        <p class="chart-subtitle">Three classifiers evaluated on 271,701 held-out test cases</p>
        <div class="chart-canvas-wrap">
          <canvas id="heroModelChart"></canvas>
        </div>
        <p class="chart-note">Higher is better for AUROC and PR-AUC. Lower is better for Brier Score. Dummy baseline uses training-set class distribution.</p>
      </div>

      <div class="chart-container">
        <h3>Monthly Arraignment Volume</h3>
        <p class="chart-subtitle">OCA-STAT records by year-month cohort, 2021&ndash;2025</p>
        <div class="chart-canvas-wrap">
          <canvas id="volumeChart"></canvas>
        </div>
      </div>
    </section>

    <!-- Strongest patterns -->
    <section class="section">
      <div class="section-header">
        <h2>Strongest Patterns</h2>
      </div>

      <div class="card-grid">
        <div class="card">
          <h3>Geography</h3>
          <p>NYC courts: ~23% conviction rate. Non-NYC courts: ~69%.
          This is the single strongest predictor in the dataset, dwarfing all other factors.</p>
        </div>
        <div class="card">
          <h3>Charge Severity</h3>
          <p>Felony cases: ~66% conviction rate. Violations: ~25%.
          Charge severity is the second-strongest signal after geography.</p>
        </div>
        <div class="card">
          <h3>County Variation</h3>
          <p>Model AUROC ranges from 0.52 (Schoharie) to 0.77 (Monroe).
          Accuracy varies substantially across New York's 61 counties.</p>
        </div>
      </div>
    </section>

    <!-- Navigate -->
    <section class="section">
      <div class="section-header">
        <h2>Explore the Analysis</h2>
      </div>

      <div class="card-grid">
        <div class="card">
          <h3>Race-Adjusted Association</h3>
          <p>Three independent methods examine whether race correlates with conviction after controlling
          for case characteristics. The Black&ndash;White gap is ~2.6 pp across all methods.</p>
          <a href="race-analysis.html" class="card-link">View race analysis</a>
        </div>
        <div class="card">
          <h3>Methodology</h3>
          <p>Scikit-learn classifiers with post-hoc calibration. 16 arraignment-time features.
          Full model specifications, metrics, and library versions.</p>
          <a href="methodology.html" class="card-link">View methods</a>
        </div>
        <div class="card">
          <h3>FAQ</h3>
          <p>What is OCA-STAT? What does the model predict? What are the limitations?
          Short answers to common questions.</p>
          <a href="faq.html" class="card-link">Read FAQ</a>
        </div>
      </div>
    </section>

    <!-- Limitations -->
    <section class="section">
      <div class="section-header">
        <h2>Limitations</h2>
      </div>
      <div class="warning-box">
        <p>
          The predictive model uses only arraignment-time information. It excludes plea negotiations,
          evidence strength, attorney quality, and criminal history detail. The pretrial analysis is
          observational, not a controlled experiment. Both branches cover New York State only (2021&ndash;2025).
          Model accuracy varies by county and demographic group.
        </p>
      </div>
    </section>

  </main>

  <!-- Footer -->
  <footer class="site-footer">
    <div class="footer-inner">
      <span>juryrig &middot; MIT License &middot; Shane Wray 2026</span>
      <span>Reference run: <code style="color: rgba(255,255,255,0.8);">20260320_231413</code></span>
    </div>
  </footer>

  <script>
    juryrigChartDefaults();

    var chartInstances = [];

    function buildCharts() {
      chartInstances.forEach(function(c) { c.destroy(); });
      chartInstances = [];
      var C = juryrigColors();

      chartInstances.push(new Chart(document.getElementById('heroModelChart'), {
        type: 'bar',
        data: {
          labels: ['AUROC', 'PR-AUC', 'Accuracy', 'Brier Score'],
          datasets: [
            { label: 'Dummy Baseline', data: [0.5, 0.4316, 0.5684, 0.2461], backgroundColor: C.muted, borderRadius: 4 },
            { label: 'Logistic Regression', data: [0.8643, 0.8364, 0.7818, 0.1474], backgroundColor: C.accent, borderRadius: 4 },
            { label: 'Hist. Gradient Boosting', data: [0.8639, 0.8314, 0.7847, 0.1499], backgroundColor: C.navy, borderRadius: 4 }
          ]
        },
        options: {
          responsive: true,
          plugins: { legend: { position: 'top' }, tooltip: { callbacks: { label: function(ctx) { return ctx.dataset.label + ': ' + ctx.parsed.y.toFixed(4); } } } },
          scales: { y: { beginAtZero: true, max: 1, ticks: { callback: function(v) { return v.toFixed(1); } } } }
        }
      }));

      var cohortData = {"2021-1":21768,"2021-2":18994,"2021-3":27062,"2021-4":28317,"2021-5":25913,"2021-6":30429,"2021-7":30727,"2021-8":33329,"2021-9":29217,"2021-10":28661,"2021-11":26224,"2021-12":24927,"2022-1":23887,"2022-2":24547,"2022-3":30541,"2022-4":29638,"2022-5":30992,"2022-6":31872,"2022-7":30280,"2022-8":31355,"2022-9":30350,"2022-10":30408,"2022-11":29952,"2022-12":27397,"2023-1":31461,"2023-2":30006,"2023-3":34958,"2023-4":33722,"2023-5":36860,"2023-6":36476,"2023-7":36076,"2023-8":38302,"2023-9":35531,"2023-10":35910,"2023-11":33615,"2023-12":30619,"2024-1":33680,"2024-2":34270,"2024-3":34868,"2024-4":34201,"2024-5":37232,"2024-6":34834,"2024-7":38143,"2024-8":36822,"2024-9":35382,"2024-10":37246,"2024-11":32161,"2024-12":30656,"2025-1":33852,"2025-2":30545,"2025-3":35037};
      var cohortLabels = Object.keys(cohortData).map(function(k) {
        var parts = k.split('-');
        return ['Jan','Feb','Mar','Apr','May','Jun','Jul','Aug','Sep','Oct','Nov','Dec'][parts[1]-1] + ' ' + parts[0];
      });

      chartInstances.push(new Chart(document.getElementById('volumeChart'), {
        type: 'line',
        data: {
          labels: cohortLabels,
          datasets: [{
            label: 'Monthly Arraignments',
            data: Object.values(cohortData),
            borderColor: C.accent,
            backgroundColor: C.accent + '1a',
            fill: true, tension: 0.3, pointRadius: 2, pointHoverRadius: 5
          }]
        },
        options: {
          responsive: true,
          plugins: { legend: { display: false }, tooltip: { callbacks: { label: function(ctx) { return ctx.parsed.y.toLocaleString() + ' cases'; } } } },
          scales: { x: { ticks: { maxTicksLimit: 12 } }, y: { beginAtZero: false, ticks: { callback: function(v) { return (v/1000).toFixed(0) + 'K'; } } } }
        }
      }));
    }

    buildCharts();
    window.juryrigRebuildCharts = function() { juryrigChartDefaults(); buildCharts(); };
  </script>
</body>
</html>