Update app.R

app.R

@@ -31,6 +31,9 @@
   # Clean up out-of-range values
   wealth_stack[wealth_stack <= 0 | wealth_stack > 1] <- NA
   
+  # Scale by 100 
+  # wealth_stack <- 100*wealth_stack
+  
   all_vals <- values(wealth_stack)
   all_vals <- all_vals[!is.na(all_vals)]
   q_breaks_legend <- quantile(all_vals, probs = seq(0, 1, 0.2), na.rm = TRUE)
@@ -417,6 +420,7 @@
               )
             )
           ),
+          
           # Time series at clicked location
           fluidRow(
             column(
@@ -430,6 +434,28 @@
                   "Tap on the map to see the IWI time-series (1990–2019) for that location.")
               )
             )
+          ),
+          
+          ## How It Works 
+          fluidRow(
+            box(
+              title = tagList(icon("cogs"), "How It Works"),
+              status = "primary", solidHeader = TRUE, collapsible = TRUE, collapsed = TRUE,
+              width = 12,
+              HTML("
+        <p>These wealth-index predictions are <strong>AI-generated</strong> by a  
+        sequence-aware neural network trained on 30 years of <em>Demographic and  
+        Health Surveys (DHS)</em> ground-truth data.</p>
+        <ul>
+          <li>🔍 57,100+ geo-referenced survey points from DHS</li>
+          <li>⚙️ Multi-spectral satellite bands & raster-to-vector feature extraction</li>
+          <li>🎯 Calibrated & validated with held-out DHS clusters (1990–2019)</li>
+        </ul>
+        <p style='font-size:12px; color:#666;'>
+          (Learn more: <a href='https://doi.org/10.24963/ijcai.2023/684'>Methodology</a>)
+        </p>
+      ")
+            )
           )
         ),