Hi All,
Seasons Greetings!
I have a question/problem as follows.
I have a pass/fail scan procedure that is performed before and after
a particular process P say. It looks like the same proportion of objects
pass/fail after P has been performed as before P. Hence is it necessary for
me to perform the scan prior to performing process P. Can I just implement
the scan after process P based on this data?
So an example with numbers:
Total no. objects = 300
1]Scan BEFORE P: #pass = 100, #fail = 200
Proportion that passed = 100/300 ~ 0.333
Scan AFTER P:
2]Out of the 100 that passed BEFORE P: #pass = 32, #fail = 68
Proportion that passed 'AFTER P' given that they passed 'BEFORE P' =
32/100 ~ 0.32
3]Out of the 100 that failed BEFORE P: #pass = 65, #fail = 135
Proportion that passed 'AFTER P' given that they passed 'BEFORE P' =
65/200 ~ 0.33
So you can see that the pass/fail proportion is v. similar for 1], 2] and 3]
above. Is it reasonable to treat each of these as a Binomial distribution
and use the approx to the Normal distribution to test if they are
statistically the same, and if they are, conclude that the pass/fail scan
rate is the same before and after performing process P?
Regards, Richard.
_________________________________________________________________
The new MSN 8: smart spam protection and 2 months FREE*
http://join.msn.com/?page=features/junkmail
|