Not really my experience when helping on, for example, gene array data analyses: there's often a lot of causal theory (or at least hardened hypotheses) developed going into the experiments. People don't just randomly measure things.