this is for holding javascript data
blasbenito edited materials_and_methods.tex
over 9 years ago
Commit id: 05eaca9b6336b9c03188fbbb66ae5f30d0fe571e
deletions | additions
diff --git a/materials_and_methods.tex b/materials_and_methods.tex
index d67e1cf..b27831b 100644
--- a/materials_and_methods.tex
+++ b/materials_and_methods.tex
...
\item Adjusted explained deviance as an intrinsic evaluation measure to assess model goodness of fit and complexity. This measure was computed for each model following this expression: 1-((cases - 1)/(cases - predictors))*(1 - ((null deviance - deviance) / null deviance)).
\end{itemize}
The leave-one-out approach was computed as
follows for each model, once per presence record available: follows:
\begin{enumerate}
\item
One model (out of 41) is selected.
\item A background radius (out of 5) is selected.
\item A test presence record
(out of 24) is selected.
\item All presences and background data within 2.5 degrees radius around the test presence are removed.
\item A GLM model (quasibinomial family) is calibrated using the input data not removed in the following step, without the test presence.
\item Adjusted explained deviation is computed from the GLM results.
\item AUC is computed using the habitat suitability values of the test presence and the pseudoabsences as inputs.
\end{enumerate}
Finally, We calibrated and evaluated 4920 models, and averaged all the adjusted explained deviance and AUC values
were averaged for each
model. combination of model and background radius. We selected all models with AUC values higher than 0.65 and adjusted explained deviance values above 0.15. We computed the average and the standard deviation of the group of best models (CITATION ENSEMBLE). The average, scaled from 0 to 1, represented habitat suitability values, while the standard deviation represented the level of agreement reached by the best models. We plotted both measures into a single map, using the \textit{whitening} method proposed by \cite{Hengl200475} to enhance visualization and model interpretation. In the final model, areas with high habitat suitability values and low standard deviation (good agreement among models) were considered robust indicators of suitable habitat for Neanderthals, while low habitat suitability values plus low standard deviation were considered good indicators of a harsh habitat, and probably, absence.
\textbf{Importance of environmental factors}