Uncertainty-Aware Model Selection with a Calibrated Probability-Generating-Function-Based Bayesian Information Criterion
- 1Queen's University
Abstract
Selecting stochastic gene-expression models from single-cell counts requires balancing goodness of fit against unnecessary mechanistic complexity. The probability-generating-function-based Bayesian information criterion (PGF-BIC) combines covariance-weighted fitting in generating-function space with a complexity penalty, allowing candidate models to be compared without reconstructing their full count distributions. However, its conventional zero-threshold rule does not account for sampling uncertainty in the fitted score difference and may therefore favor overly complex models in finite samples. To address this limitation, we develop an uncertainty-aware PGF-BIC rule that selects the more complex model only when its score advantage exceeds a data-driven threshold. We use Cantelli's one-sided inequality to motivate a selection margin expressed in terms of a standard deviation. To determine this scale, we use influence functions to quantify sensitivity to small perturbations in the data distribution and obtain a first-order description of sampling fluctuations. The resulting variance estimate accounts for variability in both the empirical probability generating function and the estimated covariance weights, yielding a data-driven threshold for assessing the complex model's score advantage. A Poisson versus Bursty benchmark shows that the calibrated rule reduces incorrect selection of the more complex model. The calibration requires neither resampling nor additional optimization, incorporating sampling uncertainty into model selection while retaining the computational efficiency of PGF-BIC.
Related
- Probability generating functions for epidemiological time series
- Correcting and georeferencing natural history occurrence records with calibrated uncertainty
- Extended amygdala function is governed by threat probability, not uncertainty
- A unified decision framework combines phenotypic and genomic information for nonlinear selection in plant breeding
- Selection vs. integration task demands shape the similarity of information neural coding
BibTeX
@misc{wang2026uncertaintyaware,
title = {{Uncertainty-Aware Model Selection with a Calibrated Probability-Generating-Function-Based Bayesian Information Criterion}},
author = {Y. Wang and Z. Shu and F. Gao and Z. Cao},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.11.750968},
url = {https://www.biorxiv.org/content/10.64898/2026.09.11.750968v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Uncertainty-Aware Model Selection with a Calibrated Probability-Generating-Function-Based Bayesian Information Criterion
AU - Y. Wang
AU - Z. Shu
AU - F. Gao
AU - Z. Cao
PY - 2026
DA - 2026/09/11
PB - bioRxiv
DO - 10.64898/2026.09.11.750968
UR - https://www.biorxiv.org/content/10.64898/2026.09.11.750968v1
AB - Selecting stochastic gene-expression models from single-cell counts requires balancing goodness of fit against unnecessary mechanistic complexity. The probability-generating-function-based Bayesian information criterion (PGF-BIC) combines covariance-weighted fitting in generating-function space with a complexity penalty, allowing candidate models to be compared without reconstructing their full count distributions. However, its conventional zero-threshold rule does not account for sampling uncertainty in the fitted score difference and may therefore favor overly complex models in finite samples. To address this limitation, we develop an uncertainty-aware PGF-BIC rule that selects the more complex model only when its score advantage exceeds a data-driven threshold. We use Cantelli's one-sided inequality to motivate a selection margin expressed in terms of a standard deviation. To determine this scale, we use influence functions to quantify sensitivity to small perturbations in the data distribution and obtain a first-order description of sampling fluctuations. The resulting variance estimate accounts for variability in both the empirical probability generating function and the estimated covariance weights, yielding a data-driven threshold for assessing the complex model's score advantage. A Poisson versus Bursty benchmark shows that the calibrated rule reduces incorrect selection of the more complex model. The calibration requires neither resampling nor additional optimization, incorporating sampling uncertainty into model selection while retaining the computational efficiency of PGF-BIC.
ER -