Primarily intended to split a factor with a non-prime number of levels
into a series of pseudo-factors, each with a prime number of levels
and which jointly define the same classes as the factor itself. The main reason to do this would be to confound one or more of the
pseudo-factors, or their interactions, with blocks using constructions
that only apply for prime numbers of levels. In this way the
experiment can be made smaller, at the cost of some treatment
contrasts being confounded with blocks.
The default method factorizes integers by a clumsy, though effective
enough way for small integers. The function is vectorized in the
sense that if a vector of integers to factorize is specified, the
object returned is a list of numeric vectors, giving the prime
divisors (including repeats) of the given integers respectively.
As with any method of factorizing integers, it may become very slow if
the prime factors are large.