Lab: Linear Models in R

Today we’re going to walk through a few faded examples of linear models to make sure you have a handle on lm fitting in R.

A Fat Model

Fist, the relationship between how lean you are and how quickly you lose fat. Implement this to get a sense ot the general workflow for analysis

library(ggplot2)
fat <- read.csv("./lm/17q04BodyFatHeatLoss Sloan and Keatinge 1973 replica.csv")

#initial visualization to determine if lm is appropriate
fat_plot <- ggplot(data=fat, aes(x=leanness, y=lossrate)) + 
  geom_point()
fat_plot

fat_mod <- lm(lossrate ~ leanness, data=fat)

#assumptions
plot(fat_mod, which=1)
plot(fat_mod, which=2)

#f-tests of model
anova(fat_mod)

#t-tests of parameters
summary(fat_mod)

#plot with line
fat_plot + 
  stat_smooth(method=lm, formula=y~x)

An Itchy Followup

For your first faded example, let’s look at the relationship between DEET and mosquito bites.

deet <- read.csv("./lm/17q24DEETMosquiteBites.csv")

deet_plot <- ggplot(data=___, aes(x=dose, y=bites)) + 
  geom_point()

deet_plot

deet_mod <- lm(bites ~ dose, data=deet)

#assumptions
plot(___, which=1)
plot(___, which=2)

#f-tests of model
anova(___)

#t-tests of parameters
summary(___)

#plot with line
deet_plot + 
  stat_smooth(method=lm, formula=y~x)

Was that relationship linear?

We might suspect that the relationship was nonlinear. Let’s see how a simple log transform works here. Note the modifications to model fitting and stat_smooth.

deet_mod_log <- lm(log(bites) ~ dose, data=deet)

#assumptions
plot(___, which=1)
plot(___, which=2)

#f-tests of model
anova(___)

#t-tests of parameters
summary(___)

#plot with line
deet_plot + 
  scale_y_continuous(trans="log") +
  stat_smooth(method=lm, formula=y~x)

Long-Lived species and Home Ranges

Do longer lived species also have larger home ranges? Let’s test this!

zoo <- read.csv("./lm/17q02ZooMortality Clubb and Mason 2003 replica.csv")

zoo_plot <- ggplot(data=___, aes(x=mortality, y=homerange)) + 
  ___()

___

zoo_mod <- lm(___, data=___)

#assumptions
plot(___, which=1)
plot(___, which=2)

#f-tests of model
anova(___)

#t-tests of parameters
summary(___)

#plot with line
zoo_plot + 
  stat_smooth(method=___, formula=___)

Nonlinear home-ranges

That definitely wasn’t linear. Look at that outlier! Let’s log our y and see how things change.

zoo_mod_log <- lm(log(___) ~ ___, ___=___)

#assumptions
___(___)
___(___)

#f-tests of model
___(___)

#t-tests of parameters
___(___)

#plot with line
zoo_plot + 
  scale_y_continuous(trans="___")+
  ___(method=___, formula=___)