workshop 2

install.packages("nlme")
library
##individual variablility known as random factor/effect/subject
##using a nested design (hieraarchial) to incoroporate individual variability (see the variability in nature)
##fixed factor = temp (cold,tepid,warm)
##random = nest
##response = mass. mass~temp+nest
##under data, as it's numbers, nest is shown as interger, not recognised as factor
eggs$nest<-as.factor(eggs$nest)
##now factor
##fully nested design, eg frog maud nested in cold, not tepid or warm. Jean is nested (exists) in warm and nowhere else
##use 1|frog|temp
install.packages("lme4")
library(lme4)
install.packages("lmerTest")
library(lmerTest)
install.packages("nlme")~
library(nlme)
install.packages("multcomp")
library(multcomp)
model<-lmer(mass~temp+(1|nest),data=eggs,REML=T) ##REML=T component means restricted maximum likelihood or nested factor included
##(1|nest tells it nest is random componenent
## (response~fixed+(1|random))
##residuals
###x causing change in y. y=mx+c < doesn't inc error, or how accurate it predicts
##residuals (points around line (regression)) residual error = points not explained well
##test residuals are normally distributed, unbiased
##unable to use levene
##test for normal distribution and that residuals are unbiased and homoscedastic
##use hist() to plot histogram
hist(residuals(model)
plot(fitted(model),residuals(model))
## can see line of best fit, teset for bias/normal/homoscedastic
##can do on fixed/categorical data, model predicts means for each group. residuals are errors around the mean
plot(model)
anova(model)
##finding out about the random factor
rand(model)
##test for differences between the random factor (egg masses between females)
##nested design over orthogonal, evaluate patterns
difflsmeans(model,testeffs="temp")
##difflsmeans(model name,testeffs="fixed factor") (test effect of fixed factors, calc means, difference of means and comparing them (similar to tukey))
with(eggs, boxplot(mass~temp))
##produces box plot with errors
posthoc<-glht(model,linfct=mcp(temp="Tukey"))
## glht(model name, multiple comparison, fixed factor
## asked to use ageneralised linear hypothesis test (GLHT)
## asked to use multiple comparison, (MCP) grouped by the fixed factor
mcs=summary(posthoc)
cld(mcs,level=0.05, decreasing=T)
##produces compact letter display, if letters match up then not significant
mean.mass<-with(eggs, tapply(mass, list (temp),mean)) ## make mean
sd.mass<-with(eggs, tapply(mass,list(temp),sd)) ## make sd
mids<-barplot(mean.mass, ylim=c(0,200), xlab = "Female", ## barplot means with error bar using sd/mean + labels
              ylab = "egg mass")
arrows(mids, mean.mass+sd.mass,mids,mean.mass-sd.mass,code=3,angle=90,length=0.1),
text(mids, mean.mass+30, labels=c("a","b","a")) ## 2 a's not sig
install.packages("vegan")