Hey, it's a boneheaded question, so please have mercy.
How does limma calculate log2 fold change from the matrix of microarray probeset intensities? I am having trouble replicating fold changes of significant genes by hand. These data are from a series of affymetrix single channel microarrays

after lmfit/ebayes:

t1 <- topTable(ebfit,coef=1,number=Inf,adjust="BH",p.value=0.05,lfc=1.5)
head(t1,n=1)
PROBEID     SYMBOL  logFC            AveExpr      t             P.Value     adj.P.Val       B
13133579    igsf21b -2.258103999    6.337697273 -14.00972646    1.90E-10    1.24E-06    13.92612694


exprs(eset['13133579',1:12])

13133579:  7.513304853  7.242395497 7.14078462  7.262557934 7.017870937 7.175894141 7.787028062 7.49573367   5.097990827 4.814967793    4.938052079 5.434358161

The control samples are 1:8 The treatment samples are 9:12 How do I calculate log2 fold change given this example? Said another way, what series of equations are used to calculate the resulting -2.25 log2 fold change for igsf21b

I hope my question is clear. I can try to elaborate further if needed. Thanks,

Normalized intensities from Affymetrix microarrays are already on a log2 scale, so one can just take differences:

> y
 [1] 7.513305 7.242395 7.140785 7.262558 7.017871 7.175894 7.787028 7.495734
 [9] 5.097991 4.814968 4.938052 5.434358
> mean(y[9:12]) - mean(y[1:8])
[1] -2.258104