Subject: Book drops Sampling

# Book drops Sampling

From: James W. Mason <masonjw>
Date: Wednesday, March 13, 1991
```At Kansas State we have two outside bookdrops, one for walk up and one
for drive up.  Both are locked anytime the library is open. This has
been the case "always." Complaints are few and most of those are because
our parking is very limited, forcing people to leave their car standing
while going inside. Of course, AV materials are never to be returned in
the drops.

While i've got everyone on the phone, we're going to do a condition
survey similar to that at Stanford and U. Illinois. statistically
speaking, how does 400 samples sound for a collection of 800,000?
thanks.

jim

Short answer: it sounds fine.

Medium answer:  For dichotomous variables (ie yes/no), a sample size
of 384 will yield a 95% confidence interval with a width of NO MORE
THAN 5 percentage points.  This is the "worst case".  If your actual
data has an increased variance (or in English, if the number of
yeses is not pretty close to equal to the number of no's), the
interval width will be even less.  This holds true for infinite
samples.  If your population is very small (so that 384 is more than
about 5% of the population) you can apply the finite population
correction to reduce the interval even further.  Does this mean that
400 specimens (the entire group is the sample, not the individual
items) is an 'adequate' sample?  Yes, if you are going to ask one
dichotomous question.  Maybe, if you are going to try to draw
elaborate correlations, do crosstabs, etc, in which case you may not
have enough observations in a particular category to draw meaningful
conclusions.  If you are looking at continuous data, then these
remarks still obtain, if The Usual Assumptions about the underlying