Greetings,
I'm relearning some of my stats and have a question on an example
given in the Box, Hunter, Hunter book.  An excellent book btw! 
Exercise 4.1 asks to construct a randomization reference distribution
from the following set of data and give the significance level:

A B B A B
3 5 5 1 8

The answer they give is 0.05, but going by how they explain to get the
significance level from distribution, it doesn't seem to me it's even
possible to get anything below 0.10.

In their example, using 11 tomato plants, 5 given mixture A and 6
given mixture B.  They end up calculating the significance level as:

154/462=0.33, where the 462 is 11!/(6!5!), the number of total
possible combinations, and the 154 is the number of actual times that
(ybarB - ybarA) is greater than the difference obtained in the 'real'
experiment.  Using the same method I get that there are only 10
possible combinations (5!/3!2!), and so the lowest significance level
obtainable is 1/10 (if the 'real' answer is the biggest possible
difference).

So, I don't see how they could get 0.05 as an answer when the lowest
possible seems to be 0.10.  Am I doing something wrong?

I know this is somewhat rambling, but if someone could point me in the
right direction I would appreciate it.
Oh, the exercise is on page 97.

Thanks for any help,
Brian
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to