Leave-one-out cross-validation (Asteraceae)

Knowledge graph centered on Leave-one-out cross-validation (Asteraceae) with 63 nodes and 123 connections. Top connected: Rocky Mountain Biological Laboratory, Gothic, Barbey's Larkspur, Nuttall's Larkspur, Salix lucida subsp. lasiandra.

Method synopsis

Iterative procedure removing single data points, refitting models, and predicting missing values to assess model accuracy using NRMSE.

Synthesized from method descriptions across 1 paper using this protocol.

Procedure from a recent peer-reviewed implementation

Steps below were extracted from the most recent peer-reviewed implementation of this protocol in the corpus — Comparing predictive measures and model functions for estimating plant biomass: lessons from a sagebrush–rabbitbrush community (2019), Plant Ecology. Implementations in other papers (listed below) may differ.

1. Model fitting and cross-validation
   Applied four model functions (simple linear, polynomial, exponential, power) to each of three predictor measures for every species. Used generalized least squares with 'gls' function in R. Applied AIC-based model selection to determine optimal variance structure. Natural log transformed data for exponential and power regressions. Used leave-one-out cross-validation (LOOCV) to generate predictions.
   Quantities: 12 models per species (4 functions × 3 predictors), 536 total regressions after excluding species with insufficient variationDuration: Not specifiedConditions: Computational analysis
   Equipment: R software, gls function, AIC model selection
2. Model accuracy assessment
   Calculated normalized root mean square error (NRMSE) using 'nrmse' function from 'hydroGOF' package. Applied correction factors for back-transformation of exponential and power models using CF = e^(SEE2/2) where SEE is standard error of estimate. Removed models with extremely low accuracy (NRMSE > 5) from dataset before analysis.
   Quantities: 517 models retained after removing 19 with NRMSE > 5Duration: Not specifiedConditions: Computational analysis
   Equipment: R hydroGOF package, nrmse function

Typical Equipment

• R software
• hydroGOF package

Output Measurements

• normalized root mean square error (NRMSE)

Papers Using This Protocol (1)

Species Studied With This Protocol

Places Where Used