Relative expression between log values

Question

[mforde84]

Hi,

I have microarray data which is Loess normalized, and I'm trying to determine how to interpret the values. If I understand correctly, the normalized values are a log ratio: log(R/G) / log(sqrt(G/R)), please correct me if I'm wrong.

I have some general questions concerning fold change estimation:

Should I be converting these values to log2 then calculating fold change by log2(group1-group2)? They already appear like normally distributed data, so I'm not sure if it's necessary or if doing so is semantically meaningful to do so? Overall how should I calculate relative expression between two groups? Say group1 and 2 have non-log2 transformed values of 4 and 8 respectively. Does that make grp2 30%+, 2x, 4x, as much signal?

Thanks for any help understanding this, Martin

[mforde84]

Since it's a log ratio could I simply convert it to base 10 then to log2? Then calculate foldchange as log2(grp1 - grp2)?