introduction to spatial point process models for distance ... · model.spde(dset, cv) g(s) cv(s); s...

10
Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance March 18, 2016

Upload: others

Post on 10-Aug-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Introduction to Spatial Point Process Models for Distance ... · model.spde(dset, cv) g(s) cv(s); s 2 make.model(fml, cv) See INLA f() model.detfun(type) log p(z)!type = "halfnormal"

Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance

Introduction to Spatial Point Process Models forDistance Sampling Data:

using the package iDistance

March 18, 2016

Page 2: Introduction to Spatial Point Process Models for Distance ... · model.spde(dset, cv) g(s) cv(s); s 2 make.model(fml, cv) See INLA f() model.detfun(type) log p(z)!type = "halfnormal"

Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance

Section

subsection

Practical 2: Introduction

Page 3: Introduction to Spatial Point Process Models for Distance ... · model.spde(dset, cv) g(s) cv(s); s 2 make.model(fml, cv) See INLA f() model.detfun(type) log p(z)!type = "halfnormal"

Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance

Section

subsection

Practical 2: Introduction

Page 4: Introduction to Spatial Point Process Models for Distance ... · model.spde(dset, cv) g(s) cv(s); s 2 make.model(fml, cv) See INLA f() model.detfun(type) log p(z)!type = "halfnormal"

Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance

Section

subsection

Practical 2: Introduction

Recall

π(Y|Λ) = exp

(|Ω| −

∫Ω

Λ(s) ds

) NY(Ω)∏i=1

Λ(si )

Our task is to design log Λ in terms of

Detection functionsAnimal/group intensityPoissibly other bits

Page 5: Introduction to Spatial Point Process Models for Distance ... · model.spde(dset, cv) g(s) cv(s); s 2 make.model(fml, cv) See INLA f() model.detfun(type) log p(z)!type = "halfnormal"

Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance

Section

subsection

Practical 2: Introduction

model log Λ

model.intercept() cmodel.fixed(cv) β · cv(s), s ∈ Ωmodel.spde(dset, cv) g(s) · cv(s), s ∈ Ωmake.model(fml, cv) See INLA f()

model.detfun(type) log p(z)→ type = "halfnormal" −1

2βz2

→ type = "exponential" −βz→ type = "hazard" log[1− exp(−( z

σ )−b)]

After designing the components, we build a joint model:

jmdl = join(model.intercept(), model.spde(dset))

Page 6: Introduction to Spatial Point Process Models for Distance ... · model.spde(dset, cv) g(s) cv(s); s 2 make.model(fml, cv) See INLA f() model.detfun(type) log p(z)!type = "halfnormal"

Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance

Section

subsection

Practical 2: Introduction

Page 7: Introduction to Spatial Point Process Models for Distance ... · model.spde(dset, cv) g(s) cv(s); s 2 make.model(fml, cv) See INLA f() model.detfun(type) log p(z)!type = "halfnormal"

Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance

Section

subsection

Practical 2: Introduction

Page 8: Introduction to Spatial Point Process Models for Distance ... · model.spde(dset, cv) g(s) cv(s); s 2 make.model(fml, cv) See INLA f() model.detfun(type) log p(z)!type = "halfnormal"

Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance

Section

subsection

Practical 2: Introduction

Page 9: Introduction to Spatial Point Process Models for Distance ... · model.spde(dset, cv) g(s) cv(s); s 2 make.model(fml, cv) See INLA f() model.detfun(type) log p(z)!type = "halfnormal"

Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance

Section

subsection

Practical 2: Introduction

The main method for running inference is idst()

idst() is a wrapper for INLA!

Argument What for?

model Put in your model here, e.g. jmdl

data Your data set, e.g. weeds

ips A data.frame of integration pointspredict A list defining what and where to predict (optional)

verbose Tell INLA to be verbose... Pass these arguments on to INLA

Page 10: Introduction to Spatial Point Process Models for Distance ... · model.spde(dset, cv) g(s) cv(s); s 2 make.model(fml, cv) See INLA f() model.detfun(type) log p(z)!type = "halfnormal"

Introduction to Spatial Point Process Models for Distance Sampling Data: using the package iDistance

Section

subsection

Practical 2: Introduction

After running idst, check out your results!

Function What for?

summary() Check if your INLA output makes senseevaluate() Obtain intensity at some locationmarginal() Posterior of a single effect or parameterplot.spatial() A spatial plot of your resultsplot.detfun() Plot the detection function