Posting introductions Go to the course discussion board on Canvas, enter the “introductions” discussion , click “Create thread”, and introduce yourself! What are you doing at NU? Please also post a photograph of yourself. This is mostly for my benefit. Otherwise, we meet on campus somewhere and I’ve no idea why you know me but I don’t know you!
Linear Model Review I This week you will focus on reviewing the concept of the linear model.
The goal of this week’s lab is just getting your feet wet, remembering how to use R, what a linear model is and getting some answers out. Like all labs, I expect you to submit a compressed RStudio project folder. You should assume any data files are in a subdirectory called “data”. Answer the numbered questions in text, referring to graphical and tabular output as needed. You may use base graphics or ggplot2, according to your preference.
This example from Gelman and Nolan 2002 Teaching Statistics: A bag of tricks, chapter 4.
# now make each figure and the corresponding r lm() output x = c(1,2,3) y = x plot(x,y,xlab="",ylab="",bty="n",xlim=c(0,3),ylim=c(0,3)) lm.1 = lm(y~x) abline(lm.1) summary(lm.1) ## Warning in summary.lm(lm.1): essentially perfect fit: summary may be ## unreliable ## ## Call: ## lm(formula = y ~ x) ## ## Residuals: ## 1 2 3 ## 0 0 0 ## ## Coefficients: ## Estimate Std.
Linear Model Review II This week you will continue to review the concept of the linear model. This week will also see the first online paper discussion.
View videos by Wednesday, 10 am.
Evaluating Assumptions.
Limits of the linear regression (not what it sounds like!).
Paper discussions board by Tuesday week 3.
Read Guthery and Bingham (2007) (See Files | Readings on Canvas for pdf).
This week will give you further practice with making plots, estimating linear models in R, and getting predicted values from those fitted models. Create a new project in RStudio named “yourlastname_lab2”. Submit the compressed project directory, which should include an R Markdown file that can be compiled directly to html without errors. The answers to the numbered questions should be in plain text. The code should be in R code chunks.
The data bleaching are in package NRES803. These data are derived from a paper on the regional severity of coral bleaching events triggered by high ocean temperatures (McWilliams et al, 2005). The first column (SST) is Sea Surface Temperature “anomalies” – that is, the difference between the average sea surface temperature for the Caribbean between Aug and October, and the long run average from 1961 to 1990. Positive values represent years in which the water was warmer than average; negative values represent cool years.
Model Selection – Week 3 In this week we’ll begin talking about model selection and inference using information theoretic methods.
Monday is Labour Day! No Class
Before Wednesday’s online session
Read MBILS chapters 3 and 4.
View:
Introduction to model selection
Likelihood
KL information
AIC differences videos
In Wednesday’s online session we will go over Exercise 1 from MBILS Chapter 4
Friday Do Lab #3.
In this week’s lab exercise we will first go through calculating the AIC values “by hand”, and then using a addon package MuMIn. Doing the calculations by hand serves three purposes: 1) you practice using R!, 2) you gain a deeper understanding of how much work the add-on package is doing for you, and 3) sometimes the add-on package won’t work for a new class of model, and if you can do it by hand you won’t be stuck.
Model Selection – Week 4 In this week we’ll look at how model selection methods lead us to the idea of multimodel inference.
Monday optional online help session 10-11, read, watch videos, post your pre-proposal.
Post pre-proposal in discussion board by Tuesday 11:59 pm, Week 5.
Before Wednesday’s online session
Read MBILS chapter 5.
View:
Multi-model inference
In Wednesday’s online session we will go over Exercises 1 & 2 from MBILS Chapter 5
We’re going to use the same dataset that you generated last week for this week’s exercise. Last week we looked at choosing a best model. This week we’re going to model average the coefficients and predictions from the models, and make “partial dependence plots”. As before, we’ll first do each step by hand, then with AICcmodavg. Below I’ve assumed you still have all the objects created in week 3 lab.