whittaker.mortalityTable: Ignore NA and 0 values without throwing up

R/whittaker.mortalityTable.R

@@ -41,20 +41,39 @@ whittaker.mortalityTable = function(table, lambda = 10, d = 2, name.postfix = ",
         table@name = paste0(table@name, name.postfix)
     }
 
+# browser()
     probs = table@deathProbs
+    orig.probs = probs
     ages = table@ages
 
     if (missing(weights) || is.null(weights)) {
         if (is.na(table@exposures) || is.null(table@exposures)) {
-            weights = table@exposures
-        } else {
             weights = rep(1, length(ages))
+        } else {
+            weights = table@exposures
         }
     }
-    # Missing values are always interpolated, i.e. assigned weight 0:
-    weights = weights * is.na(probs)
+    # Missing values are always interpolated, i.e. assigned weight 0; Similarly,
+    # ignore zero probabilities (cause problems with log)
+    weights = weights * (!is.na(probs) & (probs > 0))
+    weights[is.na(weights)] = 0
+    if (sum(probs > 0, na.rm = TRUE) < d) {
+        warning("Table '", table@name, "' does not have at least ", d, " finite, non-zero probabilities. Unable to graduate. The original probabilities will be retained.")
+        return(table)
+    }
+
+    # Normalize the weights to sum 1
+    weights = weights / sum(weights)
+
+    # We cannot pass NA to whittaker, as this will result in all-NA graduated values.
+    # However, if prob==NA, then weight was already set to 0, anyway
+    probs[is.na(probs)] = 0
+    probs.smooth = exp(whittaker.interpolate(log(probs), lambda = lambda, d = d, weights = weights))
 
-    probs.smooth = exp(whittaker.interpolate(log(probs[!NApos]), lambda = lambda, d = d, weights = weights))
+    # Do not extrapolate probabilities, so set all ages below the first and
+    # above the last raw probability to NA
+    probsToClear = (cumsum(!is.na(orig.probs)) == 0) | (rev(cumsum(rev(!is.na(orig.probs)))) == 0)
+    probs.smooth[probsToClear] = NA_real_
     table@deathProbs = probs.smooth
 
     table
@@ -64,10 +83,12 @@ whittaker.mortalityTable = function(table, lambda = 10, d = 2, name.postfix = ",
 whittaker.interpolate = function(y, lambda = 1600, d = 2, weights = rep(1, length(y))) {
     m <- length(y)
     E <- eye(m)
+    weights = weights * is.finite(y) # non-finite or missing values in y get zero weight
+    y[!is.finite(y)] = 0
     W <- diag(weights)
     D <- diff(E, lag = 1, differences = d)# - r * diff(E, lab = 1, differences = d - 1)
     B <- W + (lambda * t(D) %*% D)
-    z <- solve(B, W %*% y)
+    z <- solve(B, as.vector(W %*% y))
     return(z)
 }
 #