The structure of the initial data for modeling tree mortality using logistic regression models in the R

A variant of the description of the data structure for modeling tree mortality is given in (Kachaev, 2020).

Time series of primary measurements of tree rings are directly used to compute logistic regression models for tree mortality (Cailleret, et al., 2016).

To expand the number of variables in modeling tree mortality, we introduce derived time series: calculated as a result of tree-ring standardization methods (Bunn, 2010) and empirical mode decompositions (Donghoh and Hee-Seok, 2018).

There are currently six standardization methods available in the dplR library (Bunn, 2010). These methods are: smoothing spline - Spline, modified negative exponential curve - ModNegExp, mean - Mean, model residuals AR - Ar, Friedman smoothing - Friedman and modified Hugershoff curve - ModHugershoff. Standardized time series are inserted into the tree data structure with the addition of the "Tdetr" object: ["Spline", "ModNegExp", "Mean", "Ar", "Friedman", "ModHugershoff"].

The empirical mode decomposition method is implemented in the EMD library (Donghoh and Hee-Seok, 2018). The algorithm decomposes the original time series into a set of time series IMFn (empirical modes) plus the residual series. The total sum of the empirical modes with the residual series gives the original series. The set of time series (empirical modes with a residual series) is inserted into the tree data structure with the addition of the "Temd" object: ["imf1", "imf2", "imf3", "imf4", "res", "low "," high "]. Let's denote the original series as Series, then low = Series- (imf1 + imf2) and high = Series- (imf3 + imf4), these are the series obtained as a result of low-frequency and high-frequency filtering of the original series.

References:

Bunn A.G. (2010).  Statistical and visual crossdating in R using the dplR library." Dendrochronologia, 28(4), 251{258. ISSN 11257865. doi:10.1016/j.dendro.2009.12.001. URL http://linkinghub.elsevier.com/retrieve/pii/S1125786510000172.

Cailleret, Maxime et al. (2016), Data from: Towards a common methodology for developing logistic tree mortality models based on ring-width data, Dryad, Dataset, https://doi.org/10.5061/dryad.1bv6n

Donghoh Kim and Hee-Seok Oh (2018) EMD: Empirical Mode Decomposition and Hilbert Spectral Analysis. R package version 1.5.8.

Kachaev, Alexander (2020), “Tree ring growth data in Json format for the development of logistic tree mortality models.”, Mendeley Data, V1, doi: 10.17632/3vht95njg3.1