Skip to contents

Identify and Fit Optimal Breakpoints in Population Growth Data

fit_breakpoints.Rd

Analyzes population growth data to detect significant changes in growth patterns by:

  • Fitting piecewise linear models to log-transformed count data

  • Implementing model selection to determine optimal number of breakpoints

  • Identifying precise positions of growth pattern changes

  • Assigning data points to their respective growth segments

Usage

fit_breakpoints(
  x,
  norm = TRUE,
  n_trials = 500,
  avg_points_per_window = 3,
  available_changepoints = c(0:5),
  model_selection = "LOO",
  n_core = 4
)

Arguments

x

A bipod object containing time series data of population counts

norm

Logical indicating whether to standardize the data (mean 0, SD 1) before fitting. Standardization applies to both time and count data. Default is TRUE

n_trials

Integer controlling optimization iterations. Higher values increase likelihood of finding global optima at the cost of longer computation time. Default is 500

avg_points_per_window

Integer specifying minimum average data points required between breakpoints. Prevents overfitting by ensuring adequate data density per segment. Default is 3

available_changepoints

Integer vector defining the range of breakpoint numbers to consider during model selection. For example, c(0:5) evaluates models with 0 to 5 breakpoints. Default is 0:5

model_selection

Character string specifying model selection criterion: * "LOO": Leave-One-Out cross-validation (recommended) * "AIC": Akaike Information Criterion * "BIC": Bayesian Information Criterion Default is "LOO"

n_core

Integer specifying number of CPU cores for parallel processing. Default is 4

Value

An enhanced bipod object containing:

  • breakpoints_fit: Fitted breakpoint model

  • metadata$breakpoints: Vector of identified breakpoint positions

  • counts$group: Factor indicating segment membership for each observation

Examples

# Create a bipod object with your data
data = biPOD::sim_stochastic_exponential(
100,
c(rep(1,5), rep(0,5), rep(1,5)),
c(rep(0,5),rep(.25,10)),
15,
.1
)

# Fot your dataset
x = biPOD::init(data, "sample w.o breakpoints")
#> 
#> ── biPOD - bayesian inference for Population Dynamics ──────────────────────────
#> 
#> ℹ Using sample named: sample w.o breakpoints.
x = biPOD:::fit_breakpoints(x, n_core=1)
#> ℹ Intializing breakpoints...
#> ℹ Breakpoints optimization...
#> ℹ Choosing optimal breakpoints...
#> ℹ Median of the inferred breakpoints have been succesfully stored.

# Plot inferred breakpoints
biPOD::plot_breakpoints_posterior(x)