Coverage Equivalence Rate (CER)

• Coverage Equivalence Rate (CER) is a performance metric that measures the proportion of replicates meeting both accuracy and precision requirements in statistical and network analyses.
• It is computed using methods like Monte Carlo simulations in survey inference and stochastic geometry in ISAC networks to assess joint coverage and quality thresholds.
• CER advances traditional metrics by simultaneously evaluating interval accuracy and tightness, offering insights for practical decision-making in diverse applications.

The Coverage Equivalence Rate (CER) is a generalized performance metric that quantifies, under specified equivalence criteria, the proportion of replicates (in simulation or theoretical random draws) where an estimator or a system meets both coverage and precision requirements. Its applications span survey methodology, where it evaluates statistical inference about finite population parameters, and communication networks, where it codifies the probability of simultaneously satisfying sensing and communication quality thresholds. CER thus formalizes a stronger notion of reliability than classical coverage or error-rate metrics by jointly assessing both the accuracy and the tightness of probabilistic intervals or events around their true or desired values.

1. Formal Definitions and Mathematical Formulation

The precise definition of CER is context dependent, with distinct formulations in survey inference and integrated communication-sensing networks.

• In finite-population inference for probabilistic surveys (Dyrkton et al., 24 Nov 2025), CER is defined as the fraction of simulated confidence intervals (CIs) whose entirety lies within a narrow equivalence band centered on the true finite-population proportion. Let $R$ be the number of Monte Carlo replications, $P_k$ the true proportion in replicate $k$, $p_k$ the point estimate, and $\hat{SE}_k$ the estimated standard error. The $95\%$ Wald CI for $p_k$ is $$[L_k, U_k] = [p_k - 1.96\hat{SE}_k, p_k + 1.96\hat{SE}_k]$$. For prespecified equivalence margin $\delta$, define the indicator $I_k(\delta) = 1$ if $[L_k, U_k] \subseteq [P_k - \delta, P_k + \delta]$, and $0$ otherwise. CER is

$$CER(\delta) = \frac{1}{R} \sum_{k=1}^R I_k(\delta).$$

• In integrated sensing and communication (ISAC) networks (Gan et al., 2024), CER formalizes the probability that both the sensing and communication requirements are met. For thresholds $\epsilon_1$ (e.g., positioning accuracy via CRLB) and $\epsilon_2$ (e.g., communication SINR), the joint coverage rate is

$$CER(\epsilon_1, \epsilon_2) = P_{p \wedge c}(\epsilon_1, \epsilon_2) = \mathbb{P}\{\mathrm{CRLB} \leq \epsilon_1 \wedge \Upsilon \geq \epsilon_2\},$$

where $\Upsilon$ denotes the received SINR and CRLB is the Cramér–Rao lower bound for localization accuracy.

2. CER in Statistical Inference: Vaccine Coverage Monitoring

In the context of survey estimation and vaccine coverage studies, CER was introduced to address not only traditional coverage (i.e., the frequency with which confidence intervals cover the true proportion) but also whether these intervals are sufficiently narrow to be practically useful (Dyrkton et al., 24 Nov 2025). The procedure is as follows:

1. Each Monte Carlo replicate generates both point and interval estimates using methods such as calibration-weighted estimators or logistic regression–based imputation.
2. Empirical CIs are constructed using either Taylor-series linearization or cluster-robust sandwich variance estimation, depending on the estimator.
3. An equivalence band of width $2\delta$ (with $\delta$ commonly set to $0.05$ or $0.075$) is determined around the true replicate population parameter.
4. The interval is said to be "equivalent" if it is fully contained within this band; CER is the proportion of intervals so classified.

This approach extends the concept of interval coverage by requiring both inclusion of the truth and—crucially—sufficient precision for practical decision-making, especially in resource-constrained health monitoring settings.

3. CER in Integrated Sensing and Communication Networks

In ISAC systems, network agents must meet both sensing (e.g., localization) and communication performance. The CER metric, as defined in (Gan et al., 2024), is the joint probability of achieving both:

• Sensing coverage probability: $\mathbb{P}\{\mathrm{CRLB} \leq \epsilon_1\}$ or $\mathbb{P}\{\mathrm{SER} \leq \epsilon_3\}$.
• Communication coverage probability: $\mathbb{P}\{\Upsilon \geq \epsilon_2\}$, where $\Upsilon$ is the SINR.
• CER (joint coverage rate): The intersection probability, integrating over system random variables and topology, with closed-form double-integral, PGFL-based expressions available for Poisson cellular models.

This joint formulation addresses the coupling and trade-offs, revealing, for example, how increased base station density $\lambda$ super-linearly increases CER at low densities but exhibits diminishing gains above certain thresholds.

Parameter(s)	Survey Inference CER (Dyrkton et al., 24 Nov 2025)	ISAC Networks CER (Gan et al., 2024)
Coverage Event	95% CI within $[P_k-\delta, P_k+\delta]$	$\mathrm{CRLB} \leq \epsilon_1$ and $\Upsilon \geq \epsilon_2$
Main Inputs	$p_k$, $\hat{SE}_k$, $P_k$, $\delta$, $R$	$\epsilon_1$, $\epsilon_2$, network parameters
Formula	$CER(\delta) = R^{-1}\sum_{k=1}^R I_k(\delta)$	$$CER(\epsilon_1,\epsilon_2) = P_{p\wedge c}(\epsilon_1,\epsilon_2)$$

4. Methodological Construction and Calculation

The derivation and calculation of CER differ between statistical and network-analytic settings.

• Statistical inference pipeline:
  • For each simulated population, the estimator is applied and its CI is constructed.
  • Precision is operationalized via the equivalence margin $\delta$; only CIs that are both correct (covering $P_k$) and sufficiently precise (entirely within $[P_k-\delta, P_k+\delta]$) are counted.
  • Equivalence margins are chosen based on domain tolerances; $\delta=0.05$ and $0.075$ are common in vaccine coverage scenarios.
• ISAC network analysis:
  • CER is calculated using stochastic geometry and integral representations involving PPPs, multi-path fading, aggregate interference, and joint probability distributions of localization and SINR metrics.
  • The closed-form joint rate (CER) captures the coupling of sensing and communication due to resource sharing, inter-cell interference, and BS coordination limits.
  • Conditional ergodic rates under cross-constraints further quantify the average attainable rates when the system is guaranteed to meet one function's quality-of-service (QoS).

5. Interpretational Nuances and Relationship to Traditional Coverage

CER provides a more stringent assessment than conventional coverage rates.

• Traditional coverage quantifies the fraction of CIs including the true parameter irrespective of interval width; wide intervals with low utility may still achieve nominal coverage.
• CER penalizes both lack of coverage and excess width, reflecting a joint requirement for both statistical accuracy and utilitarian precision. For instance, a method can attain 95% traditional coverage with many overly wide CIs, yielding a low CER if few intervals are within ±$\delta$ of the true value.
• In networks, CER reflects the likelihood operations are both accurate and resource efficient, capturing critical design trade-offs.

A plausible implication is that CER should supplement, not replace, classical coverage or rate metrics, as it is sensitive to aspects of estimator behavior and system performance that conventional criteria may obscure.

6. Applications, Trade-offs, and Empirical Insights

CER has demonstrable utility in design evaluation and operational benchmark settings:

• In health survey methodology (Dyrkton et al., 24 Nov 2025), CER is a key criterion in assessing whether hybrid or anchored estimators for vaccine coverage yield practically reliable inference under non-ideal sampling and participation conditions. Realistic scenarios demonstrate that high CER (at useful equivalence margins) is achievable with low bias and acceptable traditional coverage, barring severe selection bias and low response rates.
• In ISAC networks (Gan et al., 2024), CER quantifies network reliability under simultaneous sensing and communication requirements, supporting system-level design. For example, increasing base station density from $1$ to $10$ km$^{-2}$ increases CER for stringent threshold pairs from $1.4\%$ to $39.8\%$, with further substantial gains only at very high densities.

Trade-offs are intrinsic: in networking, resource allocation to enhance either communication or sensing may degrade CER due to coupling; in survey design, narrower equivalence margins or lower sampling fractions decrease CER unless estimator variance is tightly controlled.

7. Limitations and Future Directions

CER, while capturing desirable inferential and operational stringency, is contingent upon the choice of equivalence margin, underlying estimator normality assumptions, and correct variance estimation. In networked systems, model-specific coupling effects may modulate CER sensitivity to parameters such as node density, beamwidth, and pilot budget. Extensions of CER to settings with composite or multidimensional criteria, data-dependent margins, or real-time adaptivity represent plausible avenues for further methodological development.

References

• "Anchoring Convenience Survey Samples to a Baseline Census for Vaccine Coverage Monitoring in Global Health" (Dyrkton et al., 24 Nov 2025)
• "Coverage and Rate Analysis for Integrated Sensing and Communication Networks" (Gan et al., 2024)