I would not recommend the merging of microarray data from different platforms. Neither would I recommend the merging of RNA-seq data with microarray data. These are best analysed independently, followed by a comparison of the results.
If you must do it, I would normalise each dataset independently using the standardised methods for each technology/platform, and then transform each normalised dataset into Z-scores. Then, attempt to merge, but always keeping '
ArrayVersion' as a covariate in downstream modeling.
Also take a look at my answer here: