bpdata <- read.csv("http://faculty.washington.edu/kenrice/sisg/bpdata.csv", header=T)

#additive

bpdata$n.minor <- (bpdata$snp3=="TC") + 2*(bpdata$snp3=="TT")
lm1 <- lm(sbp~n.minor, data=bpdata)
confint.default(lm1)
summary(lm1)

#dominant
bpdata$domvar <- (bpdata$snp3=="TC") | (bpdata$snp3=="TT")
lm2 <- lm(sbp~domvar, data=bpdata)
confint.default(lm2)
summary(lm2)
## some keyboards don't have |, so a trick is
bpdata$domvar <- !(!(bpdata$snp3=="TC") & !(bpdata$snp3=="TT"))

#recessive
bpdata$recvar <- (bpdata$snp3=="TT")
lm3<-lm(sbp~recvar, data=bpdata)
confint.default(lm3)
summary(lm3)

##2df
lm4<-lm(sbp~snp3, data=bpdata)
confint.default(lm4)
summary(lm4)

## question 2 ##

lm5<- lm(sbp~n.minor+dbp+bmi, data=bpdata)
summary(lm5)
# order doesn't matter for fitted model
lm5a <- lm(sbp~dbp+bmi+n.minor,data=bpdata)
summary(lm5a)
# order does matter for ANOVA table
anova(lm5)
anova(lm5a)


## question 3 ##
# doesn't work: confuses + and * with formula notation
lm6<- lm(sbp~dbp+bmi+(snp3=="TC")+2*(snp3=="TT"), data=bpdata)
# I() insulates the arithmetic computations
lm6<-lm(sbp~dbp+bmi+I( (snp3=="TC")+2*(snp3=="TT") ), data=bpdata)

## question 4 ##
#additive

glm1 <- glm(I(sbp>140)~n.minor, data=bpdata,family=binomial)
confint.default(glm1)
summary(glm1)
round(exp(confint.default(glm1)),2)

#dominant
glm2 <- glm(I(sbp>140)~domvar, data=bpdata,family=binomial)
summary(glm2)

#recessive
glm3 <- glm(I(sbp>140)~recvar, data=bpdata, family=binomial)
summary(glm3)

##2df
glm4<-glm(I(sbp>140)~snp3, data=bpdata,family=binomial)
summary(glm4)

# adjusted effect

glm5<- glm(sbp~n.minor+dbp+bmi, data=bpdata, family=binomial)
summary(lm5)