I have a question regarding the 1st level global normalization option (none vs. scaling) for fMRI analysis.

I just have a very broad understanding that it does the mean centering each volume in the time series, but wondering:

1) What's the implication of doing it (none vs. scaling)?

2) When is it preferable to use scaling, and when is it not recommended?
3) Is there a relationship between this choice and the heterogeneity of the sample?

In my case, I am working with a lifespan sample ranging from ages 18 to 79. I am wondering if it is optimal to use scaling or stick with the default option (none).

I experimented this with my dataset, and it impacted my results (2nd level) a lot (though the most significant regions seem consistent).

Any insight would be greatly appreciated.

Thank you!

Best,

Kyoungeun