Create generate_conditions_v2.R

this is an alternative of what's in generate_conditions that does the (expected?) behavior of splitting the whole domain into the boundaries indicated by the entirety of breakpoints -- treating all detected breaks as a singular hypothesis. the current implementation treats each row individually, and therefore can only split the domain into a max of four boxes at once -- the total set of fits may correspond to more than four boxes, but there is overlap in the data used.

R/generate_conditions_v2.R

@@ -0,0 +1,37 @@
+## alternative version that splits data at all combinations implied by df2, not by row
+label_conditions <- function(df1, df2) {
+  unique_x <- sort(unique(df2$x))
+  unique_y <- sort(unique(df2$y))
+
+  df1 %>%
+    rowwise() %>%
+    mutate(
+      x_label = cut(x, breaks = c(-Inf, unique_x, Inf), labels = FALSE),
+      y_label = cut(y, breaks = c(-Inf, unique_y, Inf), labels = FALSE),
+      label = paste0(
+        "x: ", case_when(
+          x_label == 1 ~ paste0("< ", unique_x[1]), ## less than min
+          x_label == length(unique_x) + 1 ~ paste0(">= ", unique_x[length(unique_x)]), ## more than max
+          TRUE ~ paste0(unique_x[x_label - 1], " <= x < ", unique_x[x_label])
+        ),
+        " & y: ", case_when(
+          y_label == 1 ~ paste0("< ", unique_y[1]),
+          y_label == length(unique_y) + 1 ~ paste0(">= ", unique_y[length(unique_y)]),
+          TRUE ~ paste0(unique_y[y_label - 1], " <= y < ", unique_y[y_label])
+        )
+      )
+    )
+}
+
+df1_labeled <- label_conditions(df1, df2)
+print(df1_labeled)
+subset(df1_labeled, x_label == 2 & y_label == 1)
+
+df2$des <- c('break detected for age 5', 'break detected for age 10')
+
+ggplot(df1_labeled, aes(x,y, pch =label)) +
+  # scale_color_grey() +
+  geom_point() +
+  # theme(legend.position = 'none') +
+  geom_hline(data = df2, aes(yintercept = y, color = des))+
+  geom_vline(data = df2, aes(xintercept = x, color = des))