Seems to work just fine here -- although globals are fraught with danger, if you're cautious the above usage doesn't seem too egregious although I don't see too much to be gained by the two steps of setting the global variable and then setting the fitoptions object variable over just setting the option itself.
>> opts=fitoptions
opts =
Normalize: 'off'
Exclude: [1×0 double]
Weights: [1×0 double]
Method: 'None'
>> global N
>> N='on';
>> opts.Normalize=N
opts =
Normalize: 'on'
Exclude: [1×0 double]
Weights: [1×0 double]
Method: 'None'
>>
