inst/doc/dormancy.R

In dormancy: Detection and Analysis of Dormant Patterns in Data

## ----include = FALSE----------------------------------------------------------
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.width = 7,
  fig.height = 5
)

## ----setup--------------------------------------------------------------------
library(dormancy)

## ----basic-example------------------------------------------------------------
set.seed(42)
n <- 500

# Create variables
x <- rnorm(n)
condition <- sample(c(0, 1), n, replace = TRUE)

# Relationship between x and y only exists when condition == 1
y <- ifelse(condition == 1, 
            0.8 * x + rnorm(n, 0, 0.3),  # Strong relationship
            rnorm(n))                      # No relationship

data <- data.frame(x = x, y = y, condition = factor(condition))

# Overall correlation is weak
cor(data$x, data$y)

## ----detect-------------------------------------------------------------------
result <- dormancy_detect(data, method = "conditional", verbose = TRUE)
print(result)

## ----conditional--------------------------------------------------------------
result_cond <- dormancy_detect(data, method = "conditional")

## ----threshold----------------------------------------------------------------
set.seed(123)
n <- 400
x <- runif(n, -2, 2)

# Relationship only emerges for extreme values of x
y <- ifelse(abs(x) > 1,
            0.9 * sign(x) + rnorm(n, 0, 0.2),
            rnorm(n, 0, 0.5))

data_thresh <- data.frame(x = x, y = y)

result_thresh <- dormancy_detect(data_thresh, method = "threshold")
print(result_thresh)

## ----phase--------------------------------------------------------------------
set.seed(456)
n <- 300
t <- seq(0, 4*pi, length.out = n)
x <- sin(t) + rnorm(n, 0, 0.2)
y <- cos(t) + rnorm(n, 0, 0.2)

# Add relationship only in certain phases
phase <- atan2(y, x)
y <- ifelse(phase > 0 & phase < pi/2,
            0.7 * x + rnorm(n, 0, 0.2),
            y)

data_phase <- data.frame(x = x, y = y)

result_phase <- dormancy_detect(data_phase, method = "phase")

## ----cascade------------------------------------------------------------------
set.seed(789)
n <- 300
x <- rnorm(n)
y <- rnorm(n)
z <- rnorm(n)

# Relationship x-y emerges when z is extreme
y <- ifelse(abs(z) > 1.5,
            0.8 * x + rnorm(n, 0, 0.2),
            y)

data_cascade <- data.frame(x = x, y = y, z = z)

result_cascade <- dormancy_detect(data_cascade, method = "cascade")

## ----trigger------------------------------------------------------------------
set.seed(42)
n <- 500
x <- rnorm(n)
z <- sample(c(0, 1), n, replace = TRUE)
y <- ifelse(z == 1, 0.8 * x + rnorm(n, 0, 0.3), rnorm(n))
data <- data.frame(x = x, y = y, z = factor(z))

result <- dormancy_detect(data, method = "conditional")

if (nrow(result$patterns) > 0) {
  triggers <- dormancy_trigger(result, sensitivity = 0.5, n_bootstrap = 50)
  print(triggers)
}

## ----depth--------------------------------------------------------------------
if (nrow(result$patterns) > 0) {
  depths <- dormancy_depth(result, method = "combined")
  print(depths)
}

## ----risk---------------------------------------------------------------------
if (nrow(result$patterns) > 0) {
  risk <- dormancy_risk(result, time_horizon = 1, risk_tolerance = 0.3)
  print(risk)
}

## ----awaken-------------------------------------------------------------------
if (nrow(result$patterns) > 0) {
  awakening <- awaken(result, pattern_id = 1, intensity = 1, n_sim = 100)
  print(awakening)
}

## ----hibernate----------------------------------------------------------------
set.seed(42)
n <- 400
time <- 1:n
x <- rnorm(n)

# Relationship that fades over time
effect_strength <- exp(-time / 150)
y <- effect_strength * 0.8 * x + (1 - effect_strength) * rnorm(n)

data_time <- data.frame(time = time, x = x, y = y)

hib <- hibernate(data_time, time_var = "time", window_size = 0.15)
print(hib)

## ----scout--------------------------------------------------------------------
set.seed(42)
n <- 300
x <- rnorm(n)
y <- rnorm(n)

# Create a region with hidden pattern
mask <- x > 1 & y > 1
y[mask] <- 0.9 * x[mask] + rnorm(sum(mask), 0, 0.1)

data_scout <- data.frame(x = x, y = y)

scout <- dormancy_scout(data_scout, grid_resolution = 15, scout_method = "density")
print(scout)

## ----plot, eval=FALSE---------------------------------------------------------
# # Overview plot
# dormancy:::plot.dormancy(result, type = "overview")
# 
# # Pattern network
# dormancy:::plot.dormancy(result, type = "patterns")
# 
# # Risk-focused
# dormancy:::plot.dormancy(result, type = "risk")
# 
# # Scout map
# dormancy:::plot.dormancy_map(scout)

## ----example_visual-----------------------------------------------------------
dormancy:::plot.dormancy_map(scout)