Mechanism vs. phenomenology in choosing functional forms: Neighborhood analyses of tree competition Case Study 3.

Slides:



Advertisements
Similar presentations
Test of (µ 1 – µ 2 ),  1 =  2, Populations Normal Test Statistic and df = n 1 + n 2 – 2 2– )1– 2 ( 2 1 )1– 1 ( 2 where ] 2 – 1 [–
Advertisements

Case Study 2 Neighborhood Models of the Allelopathic Effects of an Invasive Tree Species Gómez-Aparicio, L. and C. D. Canham Neighborhood analyses.
Analysis of variance and statistical inference.
Logistic Regression Psy 524 Ainsworth.
Likelihood Ratio, Wald, and Lagrange Multiplier (Score) Tests
Day 6 Model Selection and Multimodel Inference
Assessment. Schedule graph may be of help for selecting the best solution Best solution corresponds to a plateau before a high jump Solutions with very.
A Short Introduction to Curve Fitting and Regression by Brad Morantz
Session 2. Applied Regression -- Prof. Juran2 Outline for Session 2 More Simple Regression –Bottom Part of the Output Hypothesis Testing –Significance.
Resampling techniques Why resampling? Jacknife Cross-validation Bootstrap Examples of application of bootstrap.
Chapter 10 Simple Regression.
458 Model Uncertainty and Model Selection Fish 458, Lecture 13.
Statistics.
Chapter 11 Multiple Regression.
Quantitative Business Analysis for Decision Making Simple Linear Regression.
Hui-Hua Lee 1, Kevin R. Piner 1, Mark N. Maunder 2 Evaluation of traditional versus conditional fitting of von Bertalanffy growth functions 1 NOAA Fisheries,
Correlation and Regression
Lecture 4 Model Selection and Multimodel Inference
Statistical Methods For Engineers ChE 477 (UO Lab) Larry Baxter & Stan Harding Brigham Young University.
Inference for regression - Simple linear regression
Lecture 4 Model Formulation and Choice of Functional Forms: Translating Your Ideas into Models.
Day 7 Model Evaluation. Elements of Model evaluation l Goodness of fit l Prediction Error l Bias l Outliers and patterns in residuals.
CPE 619 Simple Linear Regression Models Aleksandar Milenković The LaCASA Laboratory Electrical and Computer Engineering Department The University of Alabama.
Simple Linear Regression Models
On Model Validation Techniques Alex Karagrigoriou University of Cyprus "Quality - Theory and Practice”, ORT Braude College of Engineering, Karmiel, May.
Foliage and Branch Biomass Prediction an allometric approach.
Forest Sampling Simulation Loukas G. Arvanitis University of Florida Robin M. Reich Colorado State University Valentina S. Boycheva Post-doctoral Associate,
The Triangle of Statistical Inference: Likelihoood
Probability Distributions and Dataset Properties Lecture 2 Likelihood Methods in Forest Ecology October 9 th – 20 th, 2006.
Mechanism vs. phenomenology in choosing functional forms: Neighborhood analyses of tree competition Case Study 3 Likelihood Methods in Ecology April 25.
Lecture 4 Model Selection and Multimodel Inference.
1 Chapter 12 Simple Linear Regression. 2 Chapter Outline  Simple Linear Regression Model  Least Squares Method  Coefficient of Determination  Model.
Analysis of Categorical and Ordinal Data: Binomial and Logistic Regression Lecture 6.
Inference for Regression Simple Linear Regression IPS Chapter 10.1 © 2009 W.H. Freeman and Company.
Review of Building Multiple Regression Models Generalization of univariate linear regression models. One unit of data with a value of dependent variable.
PCB 3043L - General Ecology Data Analysis. OUTLINE Organizing an ecological study Basic sampling terminology Statistical analysis of data –Why use statistics?
The Triangle of Statistical Inference: Likelihoood Data Scientific Model Probability Model Inference.
Likelihood Methods in Ecology November 16 th – 20 th, 2009 Millbrook, NY Instructors: Charles Canham and María Uriarte Teaching Assistant Liza Comita.
Statistical Methods II&III: Confidence Intervals ChE 477 (UO Lab) Lecture 5 Larry Baxter, William Hecker, & Ron Terry Brigham Young University.
Lecture 5 Model Evaluation. Elements of Model evaluation l Goodness of fit l Prediction Error l Bias l Outliers and patterns in residuals.
ITEC6310 Research Methods in Information Technology Instructor: Prof. Z. Yang Course Website: c6310.htm Office:
Three Frameworks for Statistical Analysis. Sample Design Forest, N=6 Field, N=4 Count ant nests per quadrat.
Seminar 3 Data requirements, limitations, and challenges: Inverse modeling of seed and seedling dispersal Likelihood Methods in Forest Ecology October.
Semi-mechanistic modelling in Nonlinear Regression: a case study by Katarina Domijan 1, Murray Jorgensen 2 and Jeff Reid 3 1 AgResearch Ruakura 2 University.
Information criteria What function fits best? The more free parameters a model has the higher will be R 2. The more parsimonious a model is the lesser.
Jump to first page Inferring Sample Findings to the Population and Testing for Differences.
Nonlinear Logistic Regression of Susceptibility to Windthrow Seminar 7 Likelihood Methods in Forest Ecology October 9 th – 20 th, 2006.
Biostatistics Regression and Correlation Methods Class #10 April 4, 2000.
Lecture 7: Bivariate Statistics. 2 Properties of Standard Deviation Variance is just the square of the S.D. If a constant is added to all scores, it has.
MEGN 537 – Probabilistic Biomechanics Ch.5 – Determining Distributions and Parameters from Observed Data Anthony J Petrella, PhD.
Functional Traits and Niche-based tree community assembly in an Amazonian Forest Kraft et al
Model Comparison. Assessing alternative models We don’t ask “Is the model right or wrong?” We ask “Do the data support a model more than a competing model?”
Linear model. a type of regression analyses statistical method – both the response variable (Y) and the explanatory variable (X) are continuous variables.
Leah Rathbun PhD Candidate, University of British Columbia
Seminar 4 - Neighborhood competition
and Other Related Measurements
Lecture 5 Model Evaluation
Spatial statistics: Spatial Autocorrelation
Lecture 4 Model Selection and Multimodel Inference
Seminar 3 - Inverse modeling
Case Study 2 - Neighborhood competition
PCB 3043L - General Ecology Data Analysis.
CJT 765: Structural Equation Modeling
Likelihood Ratio, Wald, and Lagrange Multiplier (Score) Tests
Case Study - Neighborhood Models of Allelopathy
Model Comparison.
Lecture 4 Model Selection and Multimodel Inference
Lecture 5 Model Evaluation
Lecture 4 Model Selection and Multimodel Inference
Testing Causal Hypotheses
Presentation transcript:

Mechanism vs. phenomenology in choosing functional forms: Neighborhood analyses of tree competition Case Study 3

Key References Canham, C. D., P. T. LePage, and K. D. Coates A neighborhood analysis of canopy tree competition: effects of shading versus crowding. Canadian Journal of Forest Research 34: Uriarte, M, C. D. Canham, J. Thompson, and J. K. Zimmerman A maximum- likelihood, neighborhood analysis of tree growth and survival in a tropical forest. Ecological Monographs 74: Canham, C. D., M. Papaik, M. Uriarte, W. McWilliams, J. C. Jenkins, and M. Twery Neighborhood analyses of canopy tree competition along environmental gradients in New England forests. Ecological Applications 16: Coates, K. D., C. D. Canham, and P. T. LePage Above versus belowground competitive effects and responses of a guild of temperate tree species. Journal of Ecology 97:

The general approach… where “Size” and “Competition” are multipliers (0-1) that reduce “Maximum Potential Growth”… Should these terms be additive or multiplicative? Why use 0-1 scalars as multipliers? What is “maximum potential growth”? Should we have included a “site” term?

Effect of Tree Size (DBH) on Potential Growth Lognormal function, where: X 0 = DBH at maximum potential growth X b = variance parameter Why use this function?

Recourse to macroecology? Russo, S. E., S. K. Wiser, and D. A. Coomes Growth-size scaling relationships of woody plant species differ from predictions of the Metabolic Ecology Model. Ecology Letters 10: Corrigendum: Ecology Letters 11: (deals with support intervals) Enquist et al. (1999) have argued from basic principles (assumptions) that But trees don’t appear to fit the theory…

Separating competition into effects and responses… l In operational terms, it is common to separate competition into (sensu Deborah Goldberg) - Competitive “effects” : some measure of the aggregate “effect” of neighbors (i.e. degree of reduction in resource availability, amount of shade cast) - Competitive “responses”: the degree to which performance of the target tree is reduced given the competitive effects of neighbors…

Separating shading from crowding l Most neighborhood competition studies can not isolate the effects of aboveground vs. belowground competition l The study in BC was an exception - Shading by canopy trees is very predictable given the locations, sizes, and species of neighbors (Canham et al. 1999)

Shading of Target Trees by Neighbors (as a function of distance and DBH)

Crowding “Effect”: A Neighborhood Competition Index (NCI) For j = 1 to n individuals of i = 1 to s species within a fixed search radius allowed by the plot size i = per capita competition coefficient for species i (scaled to = 1 for the species with strongest competitive effect) A simple size and distance dependent index of competitive effect: NOTE: NCI is scaled to = 1 for the most crowded neighborhood observed for a given target tree species

What if all the neighbors are on one side of the target tree? l The “Sweep” Index: - The fraction of the effective neighborhood circumference obstructed by neighbors rooted within the neighborhood l Zar’s (1974) Index of Angular Dispersion target tree

Index of Angular Dispersion (Zar 1974) where  is the angle from the target tree to the ith neighbor.  ranges from 0 when the neighbors are uniformly distributed to 1 when they are tightly clumped.

Basic Model plus Effects of Angular Dispersion  = index of angular dispersion of competitors around the target tree Bottom line: angular dispersion didn’t improve fit in early tests, so was abandoned (too much computation time)

Competitive “Response”: Relationship Between NCI and Growth

Effect of target tree size on sensitivity to competition

Sampling Considerations: Avoiding A Censored Sample… Potential neighborhood “Target” tree What happens if you use trees near the edge of the plot as “targets” (observations)?

The importance of stratifying sampling across a range of neighborhood conditions

Effect of Site Quality on Potential Growth l Alternate hypotheses from niche theory: - Fundmental niche differentiation (Gleason, Curtis, and Whittaker ): species have optimal growth (fundamental niches) at different locations along environmental gradients - Shifting competitive hierarchy (Keddy): all species have optimal growth at the resource-rich end of a gradient, their realized niches reflect competitive displacement to sub-optimal ends of the gradient Canham, C. D., M. Papaik, M. Uriarte, W. McWilliams, J. C. Jenkins, and M. Twery Neighborhood analyses of canopy tree competition along environmental gradients in New England forests. Ecological Applications 16:

Radial growth = Maximum growth * size effect * shading*crowding The full model (for any given species)... Where: MaxRG is the estimated, maximum potential radial growth DBH t is the size of the target tree, and X o and X b are estimated parameters Shading is the calculated reduction in incident radiation by neighbors, and S is an estimated parameter DBH ij and dist ij are the size and distance to neighboring tree j of species group i, and C, i and  are estimated parameters

A sample of basic questions addressed by the analyses l Do different species of competitors have distinctly different effects? l How do neighbor size and distance affect degree of crowding? l Are there thresholds in the effects of competition? l Does sensitivity to competition vary with target tree size? l What is the underlying relationship between potential growth and tree size (i.e. in the absence of competition)?

Parameter Estimation and Comparison of Alternate Models l Maximum likelihood parameters estimated using simulated annealing (a global optimization procedure) l Start with a “full” model, then successively simplify the model by dropping terms l Compare alternate models using Akaike’s Information Criterion, corrected for small sample size (AIC corr ), and accept simpler models if they don’t produce a significant drop in information. - i.e. do species differ in competitive effects? »compare a model with separate λ coefficients with a simpler model in which all λ are fixed at a value of 1

PDF and Error Distribution In our earlier study (Canham et al. 2004), residuals were approximately normal, but variance was not homogeneous (it appeared to increase as a function of the mean predicted growth)... But with a larger dataset and more higher R2, residuals were normally distributed with a constant variance…

Neutral vs. Niche Theory: are neighbors equivalent in their competitive effects? AIC corr of alternate neighborhood competition models for growth of 9 tree species in the interior cedar-hemlock forests of north central British Columbia

How do neighbor size and distance affect degree of crowding? l Both α and  varied widely depending on target tree species l  ranged from near zero to > 3 - So, depending on the species of target tree, crowding effects of neighbors ranged from proportional to simply the density of neighbors (regardless of size:  = 0; Aspen), to only the very large trees having an effect (  = 3.4, Subalpine fir) Should  and  vary, in principle, depending on the identity of the neighbor?

Does the size of the target tree affect its sensitivity to crowding? l Models including  were more likely for 5 of the 9 species: l Values for conifers were negative (larger trees less sensitive to crowding), but values for 2 of the deciduous trees were positive! Are positive values of  biologically realistic? Are the  parameter estimates “robust”? Astrup et al. 2008, Forest Ecol. Management 10: