I think the main problem with trying to compare the relative safety of flying twins verses singles is that whilst the machinery is more or less a predictable constant (i.e. one PA28r-200 will fly very much like another, as will one Seneca III fly much like another) the quality of pilots differs hugely.
Whilst I can't claim to have analysed the statistics in any great detail, save for reading all the AAIB reports that have come my way over the last 10 years, the impression that I have is the twins tend to have different accidents to singles, and the more popular the model, the more accidents it will tend to have.
Hence, if you were a spaceman coming to the planet earth and trying to decide which SEP to fly based on the sheer number of accidents, you might decide to avoid Piper Warriors in favour of BD-5's, on the basis that Warriors had more accidents...
There are huge variables in mission profile between singles and twins, with the bulk of the hours in singles being taken up with training and short flights with lots of landings in the hands of the inexperienced, verses twins being flown longer distances very often in instrument conditions, but the biggest variable is in the quality of the nut behind the wheel - i.e. the pilot.
If someone can come up with a way of putting the same quality of pilots, with the same levels of experience, talent and qualifications in both singles and twins and then compare the accident rates, you'll have a meaningful study.
Until then, you'll merely have a bunch of people using statistics as a drunk uses a lamp-post; for support, not illumination.