One of my areas of thought/study/training is inferential statistics.

Inferential statistics takes data collected on a sample of some population, analyzes and then INFERS the findings to a population. We measure samples, because MOST useful populations are far too big to measure in their entirety.

We collect data from our sample, then compare the differences in two mean values (two averages) to the variation in sample to estimate the probability that a difference of that size occurred strictly by luck and NOT due to a real difference between the two mean scores.

This, my friend is an exercise of faith, posing as science.

Do what?

I said, that the use of inferential stats is an exercise of faith. It is important because a great many sciences, those that deal with quantifiable data, depend upon inferential stats.

So where's the faith? Faith lies herein:

1) The sample is a truly representative sample of the population of interest.

2) The data are correctly measured with sufficient precision (sometimes this is NOT true!).

3) The probability outcome does NOT rule out the possibility that something happened by chance, even if the probability is quite small.

So a scientist designs research, selects a sample and then runs statistics on the data derived from their particular, and sometimes small, sample. They sometimes rely on others to input the data and run the statistical programs for them. Then the scientist takes the small probability and assumes that their judgment of the probabilities is correct.

But, my science friends aver, science requires replications. Well, a dirty secret of science is that science isn't useful until it is published. And guess what type of science tends to be VERY HARD to publish? Yes, replications of others' research studies. In fact, in a recent effort to replicate Psychological research, out of 100 studies replicated, the same results were found less than half the time!

So, if you put your total faith in science... you don't have the intellect to be doing the work.

That's my view, and I am sticking with it.

Inferential statistics takes data collected on a sample of some population, analyzes and then INFERS the findings to a population. We measure samples, because MOST useful populations are far too big to measure in their entirety.

We collect data from our sample, then compare the differences in two mean values (two averages) to the variation in sample to estimate the probability that a difference of that size occurred strictly by luck and NOT due to a real difference between the two mean scores.

This, my friend is an exercise of faith, posing as science.

Do what?

I said, that the use of inferential stats is an exercise of faith. It is important because a great many sciences, those that deal with quantifiable data, depend upon inferential stats.

So where's the faith? Faith lies herein:

1) The sample is a truly representative sample of the population of interest.

2) The data are correctly measured with sufficient precision (sometimes this is NOT true!).

3) The probability outcome does NOT rule out the possibility that something happened by chance, even if the probability is quite small.

So a scientist designs research, selects a sample and then runs statistics on the data derived from their particular, and sometimes small, sample. They sometimes rely on others to input the data and run the statistical programs for them. Then the scientist takes the small probability and assumes that their judgment of the probabilities is correct.

But, my science friends aver, science requires replications. Well, a dirty secret of science is that science isn't useful until it is published. And guess what type of science tends to be VERY HARD to publish? Yes, replications of others' research studies. In fact, in a recent effort to replicate Psychological research, out of 100 studies replicated, the same results were found less than half the time!

So, if you put your total faith in science... you don't have the intellect to be doing the work.

That's my view, and I am sticking with it.

You would have to cage us humans up like monkey's to get good data...Don't think you can ask us...that's my view and I'm sticking to it...:)

ReplyDelete