Sensitivity of Convergent Cross Mapping to temporal discontinuities: A case study of seasonal influenza in Hong Kong.
CCM is highly sensitive to data gaps, making it less reliable for incomplete time series analysis compared to GLM and PCMCI<sup>+</sup>.
Where it sits
this study against the rest of the adipotide (prohibitin-tp01) corpusSummary and findings
The study evaluated the sensitivity of Convergent Cross Mapping (CCM) to temporal discontinuities in seasonal influenza data from Hong Kong. Various data-exclusion scenarios were tested, revealing that CCM's causal inferences are highly sensitive to data gaps. In contrast, GLM and PCMCI<sup>+</sup> frameworks showed stable estimates, indicating a negative association between ozone and influenza transmission.
Abstract
Convergent cross mapping (CCM) method has been widely applied to investigate environmental drivers of infectious disease dynamics, particularly for seasonal influenza. However, its robustness to temporal gaps and missing observations-common features in surveillance data-remains largely unexplored. Using seasonal influenza surveillance data from Hong Kong, we systematically assessed the sensitivity of inferred environment-disease relationships to different data-exclusion scenarios, including the removal of low-activity periods and targeted time points. We compared CCM with quasi-binomial generalized linear models (GLM) and the Peter-Clark Momentary Conditional Independence (PCMCI<sup>+</sup>) framework. Across all scenarios, CCM-based inference exhibited pronounced sensitivity to data gaps, with both causal strength and inferred relationships varying substantially across gap configurations. In contrast, GLM and PCMCI<sup>+</sup> estimates remained stable, consistently indicating a negative association between ozone and influenza transmission. These findings highlight a critical limitation of CCM when applied to incomplete time series and underscore the need for caution in interpreting causality from gap-affected epidemiological data.
Background
This study addresses the robustness of the Convergent Cross Mapping (CCM) method in analyzing environmental drivers of infectious diseases, specifically seasonal influenza. CCM has been widely used but its sensitivity to data gaps, common in surveillance data, is not well understood. Understanding these limitations is crucial for accurate epidemiological modeling and inference.
Methods
The study utilized seasonal influenza surveillance data from Hong Kong to assess the sensitivity of CCM to temporal discontinuities. Various data-exclusion scenarios were implemented, including the removal of low-activity periods and specific time points. Comparisons were made with quasi-binomial generalized linear models (GLM) and the PCMCI<sup>+</sup> framework.
Results
CCM-based inference showed significant sensitivity to data gaps, with causal strength and relationships varying across different configurations. In contrast, GLM and PCMCI<sup>+</sup> provided stable estimates, consistently indicating a negative association between ozone levels and influenza transmission.
Interpretation
The findings suggest that while CCM can be a powerful tool, its application to incomplete time series data can lead to unreliable causal inferences. The stability of GLM and PCMCI<sup>+</sup> in this context suggests they may be more reliable for analyzing such data. This study emphasizes the importance of selecting appropriate methods for epidemiological analysis, particularly when dealing with incomplete datasets.
Key findings
- CCM showed pronounced sensitivity to data gaps.
- GLM and PCMCI<sup>+</sup> estimates were stable across scenarios.
- Negative association between ozone and influenza transmission was consistent in GLM and PCMCI<sup>+</sup>.
- CCM's causal strength varied substantially with gap configurations.
Limitations
- Sensitivity of CCM to data gaps.
- Reliance on simulated data-exclusion scenarios.
- No clinical outcomes reported.