You need to see how different sizings affect the primary
and secondary attributes I mention, and pick what does
best for your priorities.
For example, a very fast / hard-driven tgate will put
larger charge pulses onto input and output which might
be unsuitable for a sample / hold application or an ADC
input where glitch energy might end up in the data.
If your N switch and your P switch are differently
sized then their drive might be as well, and again this
might differ in turn by whether you care more about
raw speed or about clean quiet switch transitions.
Choose your size by setting up simulation test
benches that exercise the "care-abouts" and run
the same core circuit with Nwidth and Pwidth
parametric analysis, identify the relation of these
to each performance param, and in the end pick
your sweet spot overall (or by priority).