It's apples and oranges really. I would say that it's easier to use recursive descent to parse a full programming language though.

Shunting yard works very well if you want to parse math expressions into postfix notation. It is then really easy to run postfix notation into bytecode on a stack-based virtual machine.

For comparison, here is a shunting yard based math expression evaluator in Javascript that I wrote before writing TinyExpr: https://github.com/codeplea/rimath