Evaluating the BYM Model for Ecological Studies
Author Information
Author(s): Latouche Aurélien, Guihenneuc-Jouyaux Chantal, Girard Claire, Hémon Denis
Primary Institution: Inserm, U754, Villejuif, France
Hypothesis
Does introducing unnecessary spatial structure in the BYM model bias the estimation of ecological associations?
Conclusion
The BYM model does not significantly underestimate ecological associations when no spatial variation exists, except under strong autocorrelation conditions.
Supporting Evidence
- The BYM model performed well in estimating ecological associations when no spatial variability was present.
- High autocorrelation in covariates can lead to biased estimates in the BYM model.
- Models accounting for structured and unstructured residuals do not materially underestimate ecological associations.
Takeaway
This study looked at how adding extra complexity to a model can sometimes make it less accurate, especially when the data doesn't actually need it.
Methodology
A simulation study was conducted to assess the impact of unnecessary spatial structure in the BYM model on ecological association estimates.
Potential Biases
Potential overfitting due to unnecessary spatial structure could lead to biased estimates.
Limitations
The study primarily focused on specific autocorrelation structures and may not generalize to all ecological contexts.
Participant Demographics
The study analyzed data from 94 Departements of mainland France, focusing on childhood leukaemia cases.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website