Wildlife Ecology Modeling

Class Stuff

Instructor: Brian Gerber
Classroom: NR 243
- Lecture/Lab/Discussion
When: Tu 11am - 12:15pm and Fr 10am - 1pm

My Office: 202A Wagar, Colorado Cooperative Research Unit
- Office hours: TU 1:30pm - 2:30pm and by appointment
- brian.gerber@colostate.edu

Computers: Bring laptop to class
- Software/R Packages should be installed prior to lab (posted on website)
Registration:
- 680A4 001 (CRN 77149) and 680A4 L01 (CRN 77150)

What is this course?

A mix of…

statistics
modeling
coding
math / notation
science philosophy
wildlife ecology and conservation

Assessment

Assessment Components	Percentage of Grade
Course Engagement	10%
Lab Assignments	40%
Discussions	10%
Quizzes	10%
Project	30%

Project

independent or group research project - highlighting a modeling application, data/code transparency, and communication of results
group development of a lecture and lab case-study that showcases a statistical application relevant to wildlife ecology and conservation

Course Learning Objectives

Upon successful completion of this course students will be able to:

think ‘statistically’
read quantitative ecology literature
write code to fit and interpret complex statistical models relevant to wildlife ecology and conservation
communicate statistical approaches and results

Why is this class useful?

Able to read modern ecological literature
Understand what you are doing when using data and models; coding/statistics/modeling are related but not the same
Statistical modeling and coding skills are highly marketable
Taking control of your analyses
Collaborate with colleagues/statisticians

Software

Why learn to code?

efficiency
transparency
flexibility in application
shareable
marketable skill
needed for publications

Why use R?

open-source and free
small total user base / large in ecology and statistics
find help online, e.g., stackoverflow
data management
statistics
plotting / graphics

Why use RStudio?

Makes using R easier
Projects (file mgmt)
R Shiny: Interactive online apps
R Markdown: Interactive documents
Quarto: interactive articles, websites, blog, …
Posit - Certified B corp

Type of Modeling

parametric
probabilistic
generative
inferential and/or predictive

Statistics in the Modern Age

"The theory and practice of computer-age statistics are, for the most part, a case of new wine in old bottles: The fundamental tenets of good statistical thinking have not changed, but their implementation has."
- Cox and Efron, Sci. Adv. 2017;3: e1700768.

Coding in the Modern Age

Software changes all the time
Code will become obsolete
Base R functions change slower than packages
Document/Annotate code and publish it online
File management is important; use sub-folders

A Graduate Student’s Dilemma

You need to know….

field techniques
logistics / planning
people/advisor management
institutional bureaucracy
ecological theory
wildlife mgmt principles
conservation biology principles
study design
data management

public speaking
independent and team work
graphic/visual arts
‘the literature’
the job market
how to write a manuscript/thesis
writing/sharing coding
statistical modeling
…

Learning

I do not know where you are starting

Our Aim

What model did you fit? Why? How?

Think Statistically

Know…

your objective in fitting a model
the model and its properties (not just the name)
how to interpret ALL the parameters
how the parameters are being optimized
and have justification for modeling decisions
- requires reading literature
- requires evaluating assumptions yourself

Why is statistics and ecological modeling so difficult?

My Background

Science Practice

I am a pragmatist

There are many ways to do great science

There are more ways to do meh science

Disciplines have conventions

There are foundations of scientific and statistical learning

Know the why of your decisions

Ask lots of questions to everybody all the time

Teaching Philosophy

Learning is a choice (in every movement)
An inclusive environment is paramount for learning
Communication is key
Everyone has something to teach and something to learn
Struggle is good. Solving problems leads to learning
BUT….

Website

https://bgerber123.github.io/FW680A4/index.html

Class Questions

RStudio

What does each panel do?

The language of R

Objects

A storage place for information; stored in the “Environment”

‘Attributes’ describes the structure or information of the object

The language of R

Objects

# y is an 'object' that is assigned the value 3
y = 3
y

[1] 3

The language of R

Values

numeric
integer
character
factor

Objects

vector
matrix
array
list
dataframe
S3, S4, S5, and beyond

The language of R

Functions

‘does stuff’; creates or manipulates objects

‘Arguments’ are the types of things a function is asking for; the inputs

The language of R

object = function(argument1 = input1, argument1 = input2)

object = function(input1, input2)

this = sign(x = -5)

sign(-5)

[1] -1

sign(5)

[1] 1

Some useful functions

for loops

save.this=c()
for(i in 1:10){
  save.this[i] =  1-i
}

Some useful functions

Create your own function

my.mean.func = function(x){
                           sum(x)/length(x)
                          }

my.mean.func(
             c(5,4,7,2,7,1)
             )

[1] 4.333333

Some useful functions

apply/sapply/lapply/vapply

mat= matrix(rnorm(100),nrow=10, ncol=10)

apply(mat, 2, FUN=median)

 [1] -0.22895129 -0.23543173  0.25250440 -0.05512031 -0.19921642 -0.06873026
 [7] -0.15605470  0.57359061  0.32028873  0.22507929

apply(mat, 2, FUN=function(x){
  
  length(which(x>1))/length(x)
  
})

 [1] 0.1 0.2 0.1 0.2 0.1 0.0 0.0 0.4 0.3 0.3

Code Organization

Hierarchical code organization
- code structure using indenting
- top –> bottom execution

Help! My code doesn’t work…

                cor.sp.route.cor=vector("list",n.species)
cor.sp=rep(NA,n.species)
            for(s in 1:n.species){
route=new.cov.species.long.scaled[[s]]$routeID
cor.sp[s]=cor(patch.size20.species.scaled.mat.center.route[s,],patch.count20.species.scaled.mat.center.route[s,])
    for(i in 1:nroutes){
temp1=patch.size20.species.scaled.mat.center.route[s,which(route==route.id[i])]
  temp2=patch.count20.species.scaled.mat.center.route[,][s,which(route==route.id[i])]
  if(length(temp1)>5){
  cor.sp.route.cor[[s]]=abs(c(cor.sp.route.cor[[s]],cor(temp1,temp2)))
  }}}

Better…

# Create Storage objects
  cor.sp.route.cor=vector("list",n.species)
  cor.sp=rep(NA,n.species)

#loop over species    
  for(s in 1:n.species)
    {
      route=new.cov.species.long.scaled[[s]]$routeID
                        
      cor.sp[s] = cor(patch.size20.species.scaled.mat.center.route[s,],
                      patch.count20.species.scaled.mat.center.route[s,]
                      )
        # loop over species and routes                    
        for(i in 1:nroutes)
          {
            temp1 = patch.size20.species.scaled.mat.center.route[s,which(route==route.id[i])]
            temp2 = patch.count20.species.scaled.mat.center.route[,][s,which(route==route.id[i])]
            if(length(temp1)>5){
                                cor.sp.route.cor[[s]]=abs(c(cor.sp.route.cor[[s]],cor(temp1,temp2)))
                                } #End if statement
          } #End routes loop
                       
    } #End species loop