| |
 |
|
|
Science Forum Index » Space - Consult Forum » Operative characteristic curves and sampling
Page 1 of 1
|
| Author |
Message |
| rob |
 Posted: Wed Apr 30, 2008 3:46 am |
|
|
|
Guest
|
Hello.
I have to perform some tests to determine whether changed
environmental conditions (water) affected the fish.
The question is : how many batches? and how many samples (fish) in
each batch?
The null hypothesis is that there is no pollution in the fish and I
want to be sure with alpha 0.05 and beta 0.8.
I have a vague memory of Operative characteristic curves for this kind
of problem, but cannot go further than that. I remember also that
setting the maximum allowlable number of positive results to 1 instead
of 0 lead to better results. Maybe I am totally wrong.
Could someone give me a tip?
(Do not have my books with me, I left them in Europe ...)
Thanks in advance
rob |
|
|
| Back to top |
|
| David Winsemius... |
| Posted: Mon May 05, 2008 10:24 pm |
|
|
|
Guest
|
rob <robinviaggio at (no spam) yahoo.com> wrote in news:436e9785-6a1a-4827-a19d-
275cbaab5f0d at (no spam) l28g2000prd.googlegroups.com:
Quote: Hello.
I have to perform some tests to determine whether changed
environmental conditions (water) affected the fish.
The question is : how many batches? and how many samples (fish) in
each batch?
The null hypothesis is that there is no pollution in the fish and I
want to be sure with alpha 0.05 and beta 0.8.
I have a vague memory of Operative characteristic curves for this kind
of problem, but cannot go further than that. I remember also that
setting the maximum allowlable number of positive results to 1 instead
of 0 lead to better results. Maybe I am totally wrong.
Could someone give me a tip?
(Do not have my books with me, I left them in Europe ...)
Don't see any connection with receiver operating characteristic curves.
Sounds as though you need to search on "power analysis". If the question
that needs to be addresses is presence or absence of some substance, then
it may turn out that you need to think of this as a dichotomous outcome.
For your sake (i.e., the sake of smaller numbers in your design), I hope
not. You will have more power if you don't discretize measurements.
--
DW |
|
|
| Back to top |
|
| |
|
Page 1 of 1
All times are GMT - 5 Hours
The time now is Thu Jul 24, 2008 4:31 am
|
|
