Resistance.Optimization: Primary function to optimize resistance surfaces

In wpeterman/ResistanceOptimization: Optimizes continuous resistance surfaces using CIRCUITSCAPE v4.0 or greater

Description

This is the main optimization function. Simply provide the object created from running the 'Optim.prep' function. This function will optimize each resistance surface by first implementing a constrained grid search, and then using nlm.

Usage

Resistance.Optimization(Optim.input)

Arguments

Optim.input

This is the object created from running 'Optim.prep'

Value

This function will generate numerous files. In the results directory, 6 .csv files will be created (assuming that bootstrapping was conducted). 1) All_Optimization_Results gives the model transformation parameters and AIC score for every permutation tested. It includes the constrained search parameters (shape.start and max.start) as well as the final optimized values (Opt.Shape and Opt.Max) 2) TopModel_Optimization_Parameters gives only the parameters for the transformation of each surface with the lowest AIC. 3) TopModel_Optimization_Results gives the AIC table for the optimized resistance surfaces 4) Coefficient_Table contains the linear mixed effects model parameter coefficient estimates 5) Boot_AvgRank_summary gives the average model rank, sd of model rank, avg weight, sd of weight, and upper and lower confidence intervals from the bootstrap iterations 6) Boot_ModelSelection_Freq provides the percent frequency that each resistance surface was selected as the top model

In the Plots folder, there is a PDF file for each optimized surface showing the transformed response curve.

In the Final_CS_Surfaces folder, are the .asc files and Circuitscape output for the final optimization of each surface. The cumulative resistance .asc file is also generated.

Note

It is important to remember that CIRCUITSCAPE resistances are relative. That is, if two landscapes with resistances ranging from 1 to 10 and 1 to 100 are run through CIRCUITSCAPE, the resultant pairwise resistance estimates will be perfectly correlated. This poses a challenge for optimization procedures, which often settle on extremely large values. Nonetheless, the shape and relationship of transformed resistance values, as well as their fit to the provided response data, are meaningful and comparable among tested resistance surfaces.

Author(s)

Bill Peterman

Examples

# Resistance.Optimization(Optim.input=Optim.input)

wpeterman/ResistanceOptimization documentation built on May 4, 2019, 9:48 a.m.