**0**wrote:

I'm using Limma to analyze Illumina 450k methylation data. I'm comparing the methylation (M-Values) in Obese vs Lean subjects, and I have a total of 49 arrays, representing 3 Lean subjects and 11 Obese subjects. Each subject is represented by 3-5 arrays,with the exception of one Lean subject that is represented only by one array. Is it appropriate to use the duplicateCorrelation function when some biological replicates have no technical replication? I would like to keep that array in the analysis since there are so few Lean subjects in the study.

Design setup and code:

```
head(sample_pheno,n=10)
Subject Condition
1 1 Obese
2 1 Obese
3 1 Obese
4 1 Obese
5 1 Obese
6 2 Lean
7 2 Lean
8 2 Lean
9 2 Lean
10 2 Lean
#Design setup
Condition<-factor(sample_pheno$Condition)
design<-model.matrix(~0+Condition)
colnames(design)<-levels(Condition)
head(design)
Lean Obese
1 0 1
2 0 1
3 0 1
4 0 1
5 0 1
6 1 0
#calculate correlation within subjects
corfit<-duplicateCorrelation(M_Val,design,block=sample_pheno$Subject)
fit<-lmFit(M_Val,design,block=sample_pheno$Subject,correlation=corfit$consensus.correlation)
```

**7.4k**• written 3.2 years ago by dsperley •

**0**