(Code) Wavelet-Based Detection and Wavelet-Informed Linear Modeling Framework for BOLD Signal Analysis

This dataset contains R scripts used to generate the simulation results reported in “A Wavelet-Based Detection and Wavelet-Informed Linear Modeling Framework for BOLD Signal Analysis.” The code implements a controlled single-voxel fMRI simulation framework to compare a standard General Linear Model (GLM) with a phase-restricted wavelet-based detection approach and a Wavelet-Informed GLM (W-iGLM). Simulated BOLD time series are generated by convolving event onsets with canonical or latency-shifted hemodynamic response functions (HRFs) and embedding them in structured, non-stationary noise composed of low-frequency drift, 1/f (“pink”) noise, and autoregressive (AR(1)) components. The scripts include: Simulation and condition-comparison loops assessing detection sensitivity as a function of (i) signal-to-noise ratio (SNR), (ii) hemodynamic latency mismatch, and (iii) event duration. Continuous Wavelet Transform (CWT) analysis using a complex Morlet wavelet, with statistical significance masking, cone-of-influence exclusion, and phase-based filtering. Extraction of band-limited (6–66 s) post-event wavelet power for condition discrimination. A data-driven latency estimation procedure based on wavelet power peaks, used to realign GLM regressors in the Wavelet-Informed GLM (W-iGLM). Output of summary statistics (t-values, p-values, regression coefficients, and R²) as spreadsheet files for visualization and reporting. All analyses were performed in R using publicly available packages (notably biwavelet and fmrihrf). The code is provided to ensure full reproducibility of the simulations and figures reported in the associated manuscript and may be adapted for methodological exploration of temporal uncertainty in fMRI data analysis.