Can you always turn a recursive function into an iterative one? Yes, absolutely, and the Church-Turing thesis proves it if memory serves. In lay terms, it states that what is computable by recursive functions is computable by an iterative model (such as the Turing machine) and vice versa. The thesis does not tell you precisely how to do the conversion, but it does say that it's definitely possible.
In many cases, converting a recursive function is easy. Knuth offers several techniques in "The Art of Computer Programming". And often, a thing computed recursively can be computed by a completely different approach in less time and space. The classic example of this is Fibonacci numbers or sequences thereof. You've surely met this problem in your degree plan.
On the flip side of this coin, we can certainly imagine a programming system so advanced as to treat a recursive definition of a formula as an invitation to memoize prior results, thus offering the speed benefit without the hassle of telling the computer exactly which steps to follow in the computation of a formula with a recursive definition. Dijkstra almost certainly did imagine such a system. He spent a long time trying to separate the implementation from the semantics of a programming language. Then again, his non-deterministic and multiprocessing programming languages are in a league above the practicing professional programmer.
In the final analysis, many functions are just plain easier to understand, read, and write in recursive form. Unless there's a compelling reason, you probably shouldn't (manually) convert these functions to an explicitly iterative algorithm. Your computer will handle that job correctly.
I can see one compelling reason. Suppose you've a prototype system in a super-high level language like [donning asbestos underwear] Scheme, Lisp, Haskell, OCaml, Perl, or Pascal. Suppose conditions are such that you need an implementation in C or Java. (Perhaps it's politics.) Then you could certainly have some functions written recursively but which, translated literally, would explode your runtime system. For example, infinite tail recursion is possible in Scheme, but the same idiom causes a problem for existing C environments. Another example is the use of lexically nested functions and static scope, which Pascal supports but C doesn't.
In these circumstances, you might try to overcome political resistance to the original language. You might find yourself reimplementing Lisp badly, as in Greenspun's (tongue-in-cheek) tenth law. Or you might just find a completely different approach to solution. But in any event, there is surely a way.
Best Answer
There's a simple ad hoc proof for this. Since you can build a Turing complete language using strictly iterative structures and a Turing complete language using only recursive structures, then the two are therefore equivalent.