Quantifying the lighter elements, like oxygen, using EDS tends to give a reasonably high level of uncertainty. We tend to quantify the heavier elements and estimate the oxygen content based on charge balancing. If you're trying to measure the amount of Fe to Si in a compound then you really need a well characterised standard with these elements in it. You won't be able to use mixed powders of Fe2O3 and Si as you just won't know for sure what ratio mixture you're sampling from whilst collecting EDS spectra. If you're looking at a sample where the Si is in the shell and Fe2O3 are in the core I assume you're trying to measure some sort of reaction between the interface.

You can buy EDS standards (including oxides) from EM suppliers (google "eds standards" or something like that). 

If size of your particles is a few microns or less, then true EDS quantification is nearly impossible (core/shell configuration is a tough thing for EDS).

I agree with V. Dusevich. In general, you should think how reliable these data can be. I am not totally sure whether it is possible, but you can use CASINO  (simulation program for backscattered electrons and EDS signal) in order to get an impression how much you can trust in case of core/shell particles. Too many unknown parameters (thickness of shell), shape (no flat surface, i.e. multiple stimulation), sensitivity of detector (low Z elements)...

On the other hand, you know the composition: SiO2 and Fe2O3. For what exactly you need EDS? Is there no better technique to answer your question?

I agree with Vladimir and Gert, no matter the technique, you will need standard for the elementa you want to analyse im your sample. Have you thought about electron microprobe or even WDS analysis in an SEM?

Thank you so much for answers and it seems that EDS may not be a suitable technique for this sample. i probably think of using ICP-MS as an alternative technique.