[R-sig-ME] Advice for analysis of biological data - Mixed model or NESTED-Anova?
epalmery at cns.umass.edu
Wed May 25 21:58:14 CEST 2016
I think you are on the right track. If you use function nlme, you can get
your p-values straightaway.
With lme4, you have to employ another function (Likelihood ratio test on
full and reduced models, or Wald tests with Anova in car) to extract them:
For your model coding, make sure that the biggest group is listed FIRST.
So for you:
model2=lmer(logVolume ~ Group + (1|Animal_ID/Group ), data=data, REML =
brainmodel<-nlme(logVolume ~ Group , random= ~Group/Animal_ID)
See some examples under "model specification" on this very helpful page:
Here are some nlme examples:
On Wed, May 25, 2016 at 3:32 PM, Savani Anbalagan <savani1987 at gmail.com>
> Dear all,
> I was suggested in the stack exchange.com to consult in this maling list.
> I have data from image analysis of zebrafish brain structures. I will
> discuss our data below with some analogy to make my explanation clear.
> 1. Data model: Group>Animal 1..2...3....10>Volume 1..2..3.....1000
> 2. Data model: Group>Drug treatment..1..2>Animal 1..2...3....10>Volume
> 3. I am studying axonal synapses in Brain.
> 4. I have 3 or more groups (Genotypes: Wild type, Hetero, Homozygous
> 5. Animals are sacrificied to image them.
> 6. I have 10+ animals from each group.
> 7. The number and volume of the synapses are variable.
> 8. Within the group, some animals have 300 synapses, some have 450
> 9. The volume of the synapses range from 0.2 to 50. The histrogram is
> highly skewed towards lower values. A log transformation makes it look
> 10. Some times, we also treat the different groups to a drug. So, it
> makes another level.
> 1. > (Imagine a tree with fruits of different sizes. And I am interested
> in the size of the fruits)
> 2. >(Lets say, I have trees of different species. example Indian Mango
> vs Brazilian Mango vs another Mango)
> 3. >(To collect fruits, The trees are cut. )
> 4. >(10+ trees in each groups)
> 5. >(The number of fruits vary depending on tree to tree even within
> same group. The size of the fruit varies. There are relatively too many
> small fruits).
> 6. >(Some times, fertilizers are added to tree, and then effect of fruit
> count/size is also checked)
> My questions:
> Could you please let me know,
> 1. Should I perform Nested ANOVA or Mixed model analysis?
> 2. If mixed model design, should I run the analysis on log transformed
> data or raw data? Is the distribution important for mixed model
> 3. If drug treatment is added, Is it Nested or Mixed model design?
> 4. For mixed model analysis how can I calculate p-value? Could you
> please let me know for both the cases. For experiments, without any
> And for experiments with the drug treated vs control.
> 5. These are the codes that I use to analyze my data: Could you check if
> it is correct?
> My nested anova code I use:
> logGFPVol.anova = aov(logVolume ~ Group + Error(Animal_ID/Group),
> Mixed model code:
> model2=lmer(logVolume ~ Group + (1|Animal_ID/Group ), data=data, REML =
> Please feel free to ask if I am unclear.
> Many thanks,
> *Savani Anbalagan, Ph.D*
> *Dept. of Mol. Cell Biology*
> *Weizmann Institute of Science234 Herzl St., Rehovot 76100,*
> *ISRAELPhone: +972-8934-6158*
> [[alternative HTML version deleted]]
> R-sig-mixed-models at r-project.org mailing list
Department of Biology
221 Morrill Science Center
611 North Pleasant St
Amherst MA 01003
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models