Hi, I want to perform an anova analysis for identifying differentially expressed genes. Should I scale the data using quantile normalization or median absolute deviation before looking for differentially expressed genes or should I apply anova directly on data obtained through RMA?
I am working with Affymetrix single channel microarrays.

You can apply ANOVA after RMA, since it includes background adjustment, summarisation and quantile normalisation steps (see here). For non-Affy data, something along the lines of log transformation/quantile normalisation or VST (variance stablilising transformation)/RSN (robust spline normalisation) for Illumina data could be used prior to ANOVA.

By the way you will also want to do some filtering on the data after RMA and prior to ANOVA to remove probesets of low variance. If you're working in R you will want to look at genefilter.

The RMA preprocessing step most likely already involved normalization, at least if performed by running rma via expresso or justRMA from the affy package.

RMA should be enough for normalisation. But you should do some quality control and possibly filtering of bad data before you continue. There are many Bioconductor packages that help you do that We try to offer some guidance on arrayanalysis.org.

Absolutely, it's quantile normalisation, to be precise.