Dear all, I?d like some help with analysing some count data. I am very new to R (and statistical analysis for that matter!) and have done my best to work it out on my own ? but seemed to have got stuck! I am looking at the effects of cutting hedges on invertebrates (I?m not that interested in the difference of inverts collected on the east side of a hedge compared to the west) The experimental design consists of paired cut and uncut plots established in 3 different hedgerow types. Invertebrate counts were taken from each plot (6 plots overall) four times over 4 weeks. Plot Cut state Orientation No. inverts Week 1 Week 2 Week 3 Blackthorn Cut East Cut West Uncut East Uncut West Hawthorn Cut East Cut West Uncut East Uncut West Hazel Cut East Cut West Uncut East Uncut West Table 1. Example of data. The data was very skewed and contained a fair few zero counts. I therefore decided to use glm. ##I first started with a saturated model## model1<-glm(Inverts~Plot*Cut.Uncut*orientation,quasipoisson) ##Three way interactions are removed## model2<-update(model1,~.-Plot:Cut.Uncut:orientation) ##anova tests whether the three way interaction is significant or not## anova(model1,model2,test="Chi") ##I continued to strip down the model## model3<-update(model2,~.-Plot:Cut.Uncut) anova(model3,model2,test="Chi") model4<-update(model2,~.-Plot:orientation) anova(model4,model2,test="Chi") ## I then looked to see whether just plot had an effect## model5<-update(model3,~.-Plot:orientation) model6<-update(model5,~.-Plot) anova(model6,model5,test="Chi") Plot was found to be significant. I then wanted to know where this was coming from so looked at the glht function?. Summary(glht(model5,mcp(Plot=?Tukey?))) This showed Blackthorn to be significantly different to the other two hedges which looking at a box plot seems about right. However if an interaction between Plot and Cut.Uncut variables was found how would I explore this further? The glht with Tukey specified seems to only work with one variable? Apologies if my explanation is poor, I would be more than happy to give you more information if it would help. I?m not sure if anything I?ve done if right or even if I?m on the right lines? Any help would be fantastic! Many thanks, Mary
Help with glm and glht for analysing count data
2 messages · Mary Crossland, Bert Gunter
It appears that these are primarily statistical issues and, as such, are somewhat off topic here. I suggest you post on stats.stackexchange.com instead for statistical help. Also, if you insist on posting here, post in plain text, not HTML (as requested by the posting guide, which you would do well to read and follow). Cheers, Bert Bert Gunter Genentech Nonclinical Biostatistics (650) 467-7374 "Data is not information. Information is not knowledge. And knowledge is certainly not wisdom." Clifford Stoll
On Tue, Nov 4, 2014 at 6:37 AM, Mary Crossland <marycrossland at gmail.com> wrote:
Dear all,
I?d like some help with analysing some count data. I am very new to R (and
statistical analysis for that matter!) and have done my best to work it out
on my own ? but seemed to have got stuck!
I am looking at the effects of cutting hedges on invertebrates (I?m not
that interested in the difference of inverts collected on the east side of
a hedge compared to the west)
The experimental design consists of paired cut and uncut plots established
in 3 different hedgerow types. Invertebrate counts were taken from each
plot (6 plots overall) four times over 4 weeks.
Plot
Cut state
Orientation
No. inverts
Week 1
Week 2
Week 3
Blackthorn
Cut
East
Cut
West
Uncut
East
Uncut
West
Hawthorn
Cut
East
Cut
West
Uncut
East
Uncut
West
Hazel
Cut
East
Cut
West
Uncut
East
Uncut
West
Table 1. Example of data.
The data was very skewed and contained a fair few zero counts. I therefore
decided to use glm.
##I first started with a saturated model##
model1<-glm(Inverts~Plot*Cut.Uncut*orientation,quasipoisson)
##Three way interactions are removed##
model2<-update(model1,~.-Plot:Cut.Uncut:orientation)
##anova tests whether the three way interaction is significant or not##
anova(model1,model2,test="Chi")
##I continued to strip down the model##
model3<-update(model2,~.-Plot:Cut.Uncut)
anova(model3,model2,test="Chi")
model4<-update(model2,~.-Plot:orientation)
anova(model4,model2,test="Chi")
## I then looked to see whether just plot had an effect##
model5<-update(model3,~.-Plot:orientation)
model6<-update(model5,~.-Plot)
anova(model6,model5,test="Chi")
Plot was found to be significant. I then wanted to know where this was
coming from so looked at the glht function?.
Summary(glht(model5,mcp(Plot=?Tukey?)))
This showed Blackthorn to be significantly different to the other two
hedges which looking at a box plot seems about right. However if an
interaction between Plot and Cut.Uncut variables was found how would I
explore this further? The glht with Tukey specified seems to only work with
one variable?
Apologies if my explanation is poor, I would be more than happy to give you
more information if it would help.
I?m not sure if anything I?ve done if right or even if I?m on the right
lines?
Any help would be fantastic!
Many thanks,
Mary
[[alternative HTML version deleted]]
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.