Like I said in the original post:

statistics have a habit of serving their master and representing what the authors want them to.
Its all in the questions. The results are spun or pre-determined based on what the question is and the people asking the question word it to favor themselves.

The polls in the United States for the war with Iraq were like that. The news would show polls and say that 78% of Americans are in favor of attacking Iraq to disarm them. What they wouldn't say is that the poll question was something like "If the U.N. inspection team finds undeniable proof of chemical or biological weapons would you be in favor of the U.S. attacking Iraq to disarm them?" or "If the U.N. Security Council approves the use of force would you be in favor of the United States joining a multi-national force to disarm Iraq?" or something like that. They just cut out the conditional part of the question when they post the results and say that 78% of Americans are in favor.

If they would have said "If the U.N. inspection team finds no proof of weapons of mass destruction and the U.N. Security Council does not support the use of force do you believe the United States should act unilaterally and attack Iraq regardless of world sentiment?" they might have gotten a different response.

Science experiments are often pre-determined in a similar way as well. If you believe that mixing Coca Cola and mayonaise will create a strong adhesive you can set up an experiment. Your experiment will be designed from the start to prove one thing- the mixture does form a strong adhesive or the mixtures does not form a strong adhesive.

The problem is that based on the experiment you don't know what else you may have proven or created. Perhaps the mixture fails as an adhesive so you throw it all away. But, maybe the mixture cures cancer or makes a fabulous floor polish. You don't know because it doesn't fit your original question. Your experiment was only designed to prove or disprove your question and not the hundreds or thousands of other questions the experiment might answer.