Limma Paired Analysis - Force intercept to 0
0
Entering edit mode
@moltenlight-24367
Last seen 3 days ago

I have a question regarding paired analysis in the Limma package. I am analyzing Illumina methylation arrays of paired tumor samples from two different sites. I came up with the following script to analyze the data, which I put together from old scripts I found in our lab. But there is something that bothers me, I do not understand why the design matrix has an intercept that is forced to 0. From a lecture on linear models I remembered that forcing the intercept to 0 is bad practice in many cases, but I'm not sure if I'm missing the point here? The results are substantially different between r~0+Location+Pairs and r~Location+Pairs.

# this is the factor of interest
Location <- factor(pd$Sample_Group, levels = c("M","P"))

# the individual effect, each pair is a subject
Pairs <- factor(pd$Pair)

# design matrix
design <- model.matrix(~0+Location+Pairs, data=pd)
colnames(design) <- c(levels(Tissuetype),levels(Pairs)[-1])

# fit the linear model 
fit <- lmFit(Beta_To_M(myCombatLimma), design)

# create a contrast matrix for specific comparisons
contMatrix <- makeContrasts(P-M, levels=design)

# fit the contrasts, estimate sample-variance, compute moderated t-statistics 
fit2 <- contrasts.fit(fit, contMatrix)
fit2 <- eBayes(fit2)

# pairwise analysis
summary(decideTests(fit2,
                    lfc = 0, # cutoff based on beta values
                    p.value = 0.05,
                    adjust.method = "BY",
                    method = "separate"))
limma MethylationArray • 53 views
ADD COMMENTlink
2
Entering edit mode
@gordon-smyth
Last seen 34 minutes ago
WEHI, Melbourne, Australia

The intercept isn't being forced to zero, it is just a reparametrization. You will get exactly the same results regardless of whether you specify 0+ or not if you make the corresponding change to the contrast. See Section 9.2 of the limma User's Guide.

ADD COMMENTlink

Login before adding your answer.

Similar Posts
Loading Similar Posts
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.3