This is tough. Statistical studies necessarily need to take years of data to even make a hint about causality. Controlled studies are difficult because the controls are difficult to recreate.

Traffic, once you get down to it, shows properties of emergent behaviors; behaviors which can be attributed to the sum total of traffic, despite each individual doing their own thing and knowing nothing of the behavior of the system. Therefore, it is very difficult to parse out the effect of a single aspect of the system. You can control the experiment the classic way, with leaving the object (say, bike lanes) out, and comparing traffic behavior to when they were in, but this does not necessarily form a causal relationship between bike lanes and traffic behavior because there may be a long line of behavioral modifications (which run through different parts of the system) between the object and the behavior.

Statistical studies are only as good as the data collected. Any inconsistency, any non-ideality, can be used to discredit and overturn a statistical study. Not to mention the whole emergent behavior thing which can muddle up the link between the object and the behavior.

A good example of emergent behavior running things amok is in the anti-bike lane argument when responding to reports of the experiences of people in cities where there are lots of bike lanes. The usual reponse, which the OP is guilty of giving over in another thread, is that: "well, yes, you feel safer, but are you?" It is a good question and not one which can be answered. First off, bike lanes affect both the motorist and the cyclist in just being there. We argue a lot over what this effect can be. Second, the effect of the cyclist feeling safer may lead to them actually being safer, depending on the environment surrounding the cyclist. Or not. One of the arguments commonly used about safety equipment is that it makes the user less safe because they use the increased safety margin, not to increase their safety, but to push their abilities.

Because everything is so interconnected on the road, I tend to agree with Hurst when he states that urban cycling is less a science than an art. This is why experiences of skilled cyclists cannot be counted out, and why it should have a premium over reasoning. Anyone can reason themselves to any conclusion they wish. The system is so complicated and interwoven that any line of logical argument can be used to arrive at any conclusion. Experience, though, tells all; albeit, from a narrow viewpoint and a biased one at that.

Long story short, I don't think any scientific study is going to answer the questions we have once and for all. All we can do is to try many things, and let Darwin do its work. Pay less attention to logical reasoning and more attention to the opinions of experienced cyclists. While the mind is fairly bad at logic, it excels at cutting through the complexity of emergent systems. If we gather enough experience together, and stop trying to shout each other down; if we try lots of things and quit trying to cut the legs out from under others who are also trying their best to solve the problem; if we allow bad designs to be culled and keep an open mind to allow our favorite to slip away; only then can the problem of better integrating cyclists into our traffic system be solved.

One final word: If there were a study to be done to arrive at the final answer which all can agree on, it would have been done by now. What's left is to do our best on our own corners of the world the way we think it should be done, sit back, watch the show, and let Darwin do it's thing.