Validity of using statistics Help!!! Bio Lab question?

Question

Why would you expect to hear more arguments about the validity of using statistics then the validity of using probability? This is a bio lab question and I'm stumped.

Answer 1

Usually you're dead on Sauce, but this time I have to disagree with your logic.

It's true that I usually see more arguments about the use of statistics than the use of probability. Probability is just one form of statistics after all, and it's pretty straightforward compared to some of the more complex statistical analyses. Statistics can also be predictive, and they are only as interpretive if pushed to limits. If anything, statistics is what should be used to prevent interpretation.

But the thing is, for any statistical analyses there are assumptions involved. For example when you use probability to solve mendelian genetics problems you are making some assumptions, like the two genes aren't linked, etc...

Most often these assumptions are tested before the analyses. The quality of your statistics usually goes hand in hand with the safety of your assumptions, regardless of what kind of statistics you use. Probablity can be just as tenuous as other statistical analyses if you can't justify or support your assumptions.

Answer 2

Because probabilities are predictions based on known quantities. The only difficulty is in computing all the different combinations, but this is not subject to debate. It's straight math.

Statistics, on the other hand, are measurements of unknown quantities. They are not so much predictive as interpretive, and people can debate the interpretation. For example, statistics are used to show a correlation between two variables A and B, but someone may object that there is another variable C that affects both of them. If you failed to measure C, then you can't address that objection.

For example, probabilities would be used to compute the odds that a child is colorblind based on what we know about the parents. There isn't much debate here ... if we know what the parents have, we can compute the odds of the child being colorblind, and there isn't much room for debate.

However, if you're trying to show some correlation between colorblindness and the color of a child's eyes, there are so many other variables that come into play that any statistics you use can be called into question.

Answer 3

Adding to what secretsau...said:
Statistics is based on sampling. It is also based on probability theory. In statistics you do not prove a hypothesis. You either 'reject' a hypothesis or 'fail to reject' it.
We set a low probability level for discremination, and say an event of such low probability may not be a chance occurence and hence reject it. If the occurence of the event had a higher probability than this threshold, we will 'fail to reject it' and hence accept it.