MID-TERM BIOL 300: October 2008

For all statistical tests, make sure that you clearly state your
hypotheses. Unless otherwise stated, assume α = 0.05. Show your
work. Be as precise as possible about P-values.
Some questions have a box for the final answer. Please put the final
answer in this box, and show all work in the other space provided,
including the back of the page if necessary.
By taking this test and putting your name above, you are declaring
that your answers on this test are all your own work.
Make sure that your copy of the test includes 6 pages, including this
one.

1. In order to reach its feeding area, a fish must cross a dangerous patch of water where
there is a 10% chance of being eaten by a predator, for each independent crossing. What
is the probability that a fish can survive the first 10 crossings? (10 points)
1. 0.35
The probability that it survives one cross is 1-Pr[dies per cross] = 1 - 0.1 = 0.9.
Each crossing is independent, so the probability of surviving ten times is 0.910 =
0.35.

2. Define Type I error. What is the relationship between the Type I error rate and the
significance level of a hypothesis test? (6 points)
Type I error is rejecting a true null hypothesis. The probability of a Type I error is
equal to the significance level, or α.
3. For each of the following scenarios, identify the best statistical test to use and state
the null hypothesis. (Please note, do not give the answer to the specific question, but
simply state the best test to use and the null hypothesis for the scenario.) (5 points each)
a. Comparing the mean length of left claws to mean length of right claws in a
sample of hermit crabs. (Each crab is measured for both claws.)
Paired t-test.
H0: The mean lengths of left and right claws are the same.
b. Asking whether elephant dung appears randomly distributed over space.
χ2 Goodness of fit test
H0: The number of elephant dung per square meter follows a Poisson distribution.
 2

c. Twenty bird territories are examined, and for each territory three male birds
compete to gain the territory. The study asks whether the largest male is most
likely to win these territorial battles.
Binomial test. (Goodness of fit test is also an acceptable answer. Contingency
analysis is not, because the failures and the success for each territory are not
independent, and so the individuals are not a random sample.)
H0: The largest bird succeeds one-third of the time.

d. Comparing the proportions of students who get A's, for an experiment where
one group is allowed to study and the other group is required to stay up all the
night at a party.
χ2 contingency analysis
H0: The proportion of A's is the same in the study group as the party group.

e. Asking whether there is a relationship between which drug a patient is given

and whether they survive for a month after treatment, with a very large sample.
χ2 contingency analysis
H0: Survival and drug treatment are independent.

f. Asking whether there is a relationship between which of two drugs a patient is

given and whether they survive for a month after treatment, in a small sample
where the expected number of patients who die on one of the drugs is 2.3.
Fisher's exact test (because the expected values of one of the cells are too low to
use the χ2 test.)
H0: Survival and drug treatment are independent.

4. We are often happy to do favors for other people when they have a particular need. For
example, we are more willing to let someone use a photocopier when they ask "Can I go
in front of you, because I am in a rush?" than when they give no reason: "Can I go in
front if you?"
Some researchers believe that simply giving a reason -- using the word "because" -- may
be enough to trigger this giving behavior, even when the reason is not a very good one.
An experiment was done by approaching 60 people who were about to use a photocopy
machine (Langer et al. 1978). In 30 cases (call these the "request only group", the
investigator asked, "May I use the Xerox machine?" For the request only group, 18
people allowed the investigator to go first. In the other group -- call this the "bad reason"
group-- the investigator said, "May I use the Xerox machine, because I need to use the
 3

Xerox machine?" Of the 30 people approached in this way, 28 allowed the investigator to
go first.
Do an appropriate hypothesis test to ask whether the "bad reason" approach is better or
worse than the "request only" approach. (15 points)
Here are the observed values and the expected values for a contingency table.

Observed request only group bad reason group totals

Say yes 18 28 46
Say no 12 2 14
totals 30 30 60

Expected request only group bad reason group totals

Say yes 23 23 46
Say no 7 7 14
totals 30 30 60

We'll do a χ2 contingency analysis, to test the hypotheses:

H0: People agreeing to the request was independent of whether a reason was
given.
HA: People agreeing to the request was not independent of whether a reason was
given.
(18 − 23) 2 28(−23) 2 (12 − 7) 2 (2 − 7) 2
χ2 = + + + = 9.32
23 23 7 7

There is (2-1)(2-1)= 1 degree of freedom, so the critical value is χ2(0.05).1 = 3.84.

We therefore reject the null hypothesis, and say that compliance depends on the
nature of the request.
More specifically, P < 0.005.
5. In the wake of this economic crisis, gold is getting more valuable. A new business
opens to sell gold, and they advertise that they have extraordinarily accurate scales. You
take a piece of gold bullion that you know to weigh 1.000 ounce, and have them measure
it on their scale 10 times. The answers are
1.001 1.001 1.000 1.002 0.999 0.998 1.000 0.997 0.998 0.998
a. Do a hypothesis test to ask whether the scale is accurate. (15 points)
We should do a one-sample t-test, with the hypotheses:
 4

H0: The mean value of the measurements is 1.000.

HA: The mean value of the measurements is not 1.000.

The mean and standard deviation of these measurements are

Y = 0.9994 s = 0.001647
Therefore the standard error of the mean is
s
SE Y = = 0.001647 / 10 = 0.0005207
n
Y − µ 0 0.9994 − 1.000
and t = = = −1.15 , with 10-1 = 9 degrees of freedom.
SE Y 0.0005207
This value of t is not outside the critical value of 2.26, so we do not reject the null
hypothesis. There is no strong evidence that the scale is unfair.

b. What is the median measurement of the weight of the gold bullion? (4 points)
5b. 0.9995
Putting the measurements in order, we get
0.997 0.998 0.998 0.998 0.999 1.000 1.000 1.001 1.001 1.002
There are an even number of data points, so we look at the middle two values:
0.999 and 1.000. The average of these two middle points is 0.9995.

6. Population growth per year is approximately normally distributed among countries,

with mean of 1.38% and standard deviation equal to 1.20%.
a. About what fraction of countries have a positive (greater than 0) population
growth rate? (5 points)
6a. 0.875
As population growth is approximately normally distributed, we can use the
standard normal distribution to answer this. Calculate
x − µ 0 − 1.38
Z= = = −1.15 . The probability of getting a Z > -1.15 is what we
σ 1.2
are interested in, but the table doesn’t give it. We find Pr[Z >-1.15] = 1 - Pr [Z<-
1.15] = 1-Pr [Z> 1.15]= 0.875.

b. What fraction of countries have a growth rate between -1% and 2%? (5 points)
6b. 0.6746
Pr[-1 < growth rate < 2] = 1 - Pr[ growth rate <-1] - Pr[growth rate >2].
 5

x − µ −1− 1.38
Calculate the Z values for both cases: Z = = = −1.98 and
σ 1.2
2 − µ 2 − 1.38
Z= = = 0.52 . So we need to calculate
σ 1.2
1− Pr[Z < −1.98] − Pr[Z > 0.52] = 1− 0.02385 − 0.30153 = 0.6746

7. A random sample of 200 elephants has a mean trunk length of 1.5 meters. Trunk
length is normally distributed, and 95% of the elephants in the sample have trunks
between 1.0 and 2.0 meters. Using the information from this sample, what is the 95%
confidence interval for the mean length of elephant trunks? (10 points)
Because trunk length is normally distributed, 95% of values should fall between
the mean minus 1.96 standard deviations and the mean plus 1.96 standard
deviations. So the standard deviation is approximately 0.5/1.96 = 0.255 in this
0.255
sample. Therefore the standard error is approximately SE Y = = 0.018 . Thus
200
the confidence interval for the mean is approximately
Y ± SE Y t 0.05(2),200 = 1.5 ± (0.018)(1.97) = 1.5 ± 0.036 .