In this video, I show how how to implement linear models, generalized linear models and generalized least squares models in R. Using the "airquality" dataset, I show how to fit and interpret the models. The core of the session is the interpretation of partial slope coefficients in Poisson generalized linear models. Finally, I give an outlook on generalized additive models which will be covered in one of the next sessions.
The R code used in this video is:
airquality
plot(Ozone~Wind,airquality)
model1=lm(Ozone~Wind,airquality)
plot(model1)
coef(model1)
(Intercept) Wind
96.872895 -5.550923
#predictions for Wind speeds of 19 and 20 mph:
Ozone1=coef(model1)[1]+coef(model1)[2]*19
Ozone2=coef(model1)[1]+coef(model1)[2]*20
Ozone1
Ozone2
##
model2=glm(Ozone~Wind,airquality,family=poisson)
coef(model2)
(Intercept) Wind
5.0795877 -0.1488753
Ozone1.glm=exp(coef(model2)[1]+coef(model2)[2]*19)
Ozone2.glm=exp(coef(model2)[1]+coef(model2)[2]*20)
Ozone2.glm/Ozone1.glm
0.8616765
exp(coef(model2)[2]) #exp(-0.1488753 )
#0.8616765
##
library(nlme)
model3=gls(Ozone~Wind,airquality)
summary(airquality$Ozone)
model3=gls(Ozone~Wind,airquality,na.action=na.exclude)
head(airquality)
airquality$Date=
as.Date(paste(1973,airquality$Month,airquality$Day,sep="-"))
library(lattice)
xyplot(Ozone~Date,airquality)
model4=gls(Ozone~Wind*Date,airquality,na.action=na.exclude)
air2=subset(airquality,complete.cases(Ozone))
model5=gls(Ozone~Wind*Date,air2)
plot(ACF(model5,form=~Date),alpha=0.05)
model6=update(model5,correlation=corAR1())
library(MuMIn)
AICc(model5,model6)
df AICc
model5 5 1099.40
model6 6 1095.21
summary(model6)