Omar,
I am assuming that you would be selecting a *simple random sample* from you
population of files.
Specifying the "desired error level" is entirely up to you, if I understand
your query correctly. Just as the level of confidence is entirely up to you:
in other words, are you comfortable with 90% confidence, or would you prefer
95%, or even, perhaps, 99%?
Thus, do you want your estimate of the mean number of incomplete items to be
within plus or minus 1, 2, 5, or what?
One issue is estimating the standard deviation of the number of incomplete
items. If you have no idea then, perhaps, you can estimate the range: at
least a incomplete items and at most b incomplete items, where a is the
minimum and b the maximum. Then, using the "range rule of thumb", you can
estimate the standard deviation as a-b/4 =~ std. deviation of the number of
incomplete items. (I am not a big sampling expert, so perhaps some more
sophisticated estimation techniques are available at this time. I'm just not
aware of them. Perhaps, other people on the list might be able to help.)
Remember that if the sample yielded by the original computation represents
more than 5% of the population (N=1200), then you need to adjust the
original the sample size with the *finite population correction* (fpc)
factor.
So the steps would be:
1. Estimate the "s" std. deviation of the number of incomplete items.
2. Compute the sample size (95% confidence): n = (t^2)(s^2) / (e^2); where
"t" is the value corresponding to the level of confidence (in the
neighborhood of 2 for 95%), "s" the std. deviation; "e" the error you have
decided on; "^2" squares the value.
3. Is n/N > 5% of N? If yes:
4. Apply the fpc: n * N / (N+n) = n'; where n is the original sample size; N
the population size; and n' is the final sample size (the number of files)
you will want to select.
Good luck.
Dominic Lusinchi
Statistician
Far West Research
Statistical Consulting
San Francisco, California
415-664-3032
www.farwestresearch.com
-----Original Message-----
From: SPSSX(r) Discussion [mailto:
[hidden email]] On Behalf Of
Omar Farook
Sent: Saturday, October 07, 2006 11:54 AM
To:
[hidden email]
Subject: error level
Dear Experts,
How do we specify the desired error level?
e.g., my population is 1200 file and I would like to select a sample to
monitor the number of incomplete items in each file.
Many thanks.
Omar.
---------------------------------
Get your email and more, right on the new Yahoo.com
Locked
