The secret assumption of transfer functions: problems with spatial autocorrelation in evaluating model performance

By R. Telford, Harry John Betteley Birks. 2005

282 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Atmospheric Science, Environmental Engineering

View Paper: https://www.semanticscholar.org/paper/1784b72ece7812acb5900e41279b16297dddb4d6

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Atmospheric Science

In Environmental Engineering

Citations in the corpus

Remote sensing of forest pest damage: a review and lessons learned from a Canadian perspective* (2016, background): “…to ensure the training and validation units are sufficiently far from each other as to avoid spatial autocorrelation (i.e., the tendency that sites close to each other are more similar than those farther away), which would diminish the independence of both subsets (Telford and Birks 2005).”
Is spatial autocorrelation introducing biases in the apparent accuracy of paleoclimatic reconstructions (2011, methodology): “Guiot and de Vernal (2011) comment on three of our papers, Telford and Birks (2005, 2009) and Telford (2006), that concern the implications of spatial autocorrelation for quantitative palaeoecological transfer functions.”
Understanding what controls the spatial distribution of fish populations using a multi-model approach (2011, background): “This results in apparent predictive performance being closely related to model fitting performances, and will usually lead to selection of 595 complex models with low bias and high variance, i.e. with low predictive power (Loots et al., in press, Telford and Birks, 2005).”
Spatial semiparametric models improve estimates of species abundance and distribution (2014, background): “…the model), GAMs have the ability to match virtually any pattern present in the data, causing concerns about model overfitting and the true predictive power of such models (Telford and Birks 2005; NMFS 2013), though cross-validation methods can reduce this concern (Wood 2006; Banerjee et al. 2008).”
A review of diatom δ18O in palaeoceanography (2009, background): “Furthermore, any reconstruction may also be dependent on the mathematical uncertainties associated with the use of transfer functions (e.g., Birks, 1998; Telford et al., 2004; Telford and Birks, 2005).”
Performance evaluation and hyperparameter tuning of statistical and machine-learning models using spatial data (2018, background): “Subsequently, they are aﬀected by underlying spatial autocorrelation by a varying magnitude (Brenning, 2005; Telford & Birks, 2005).”
Development of a new pan-European testate amoeba transfer function for reconstructing peatland palaeohydrology (2016, background): “Recent debates in this field have focussed on 1) more rigorous analysis of transfer function 79 results, whether via statistical testing (Telford and Birks, 2005, 2009, 2011a, 2011b, Payne et al., 80 2012, 2016; Amesbury et al., 2013), or by comparison with instrumental data (Swindles et al., 81…”
Forward selection of explanatory variables. (2008, methodology): “…of spatial analyses using principal coordinates of neighbor matrices (PCNM; Borcard and Legendre 2002, Borcard et al. 2004) and Moran’s eigenvector maps (MEM; Dray et al. 2006), which, by construction, generate a large number of variables modeling the spatial relationships among the sampling sites.”
From Classical to Canonical Ordination (2012, methodology): “To relate two communities, ter Braak and Schaffers (2004) proposed a model of co-correspondence analysis (see also Schaffers et al.”