My rule of thumb is to either do a if-return without an else (mostly for errors) or do a full if-else-if chain over all possibilities.

This way I avoid trying to be too fancy with my branching, since whenever I end up doing that my I encode application logic into my ifs, making them harder to maintain (For example, if an enum OK or ERR and I write all my ifs taking advantage of the fact that !OK <-> ERR it becomes a pain in the ass to add a third option to the enum next time.)

In your case, for example, I would use a plain if, since "return if null" is sure to be a common pattern and its not likely that you will need to worry about a third possibility (in addition to null/not null) in the future.

However, if the test was something that did a more in-dept inspection over the data I would err towards a full if-else instead.