ST1232
Tutorial 6
The resilience of genetically modified plants to a particular disease was examined in an
experiment. There are two types of genetic modification being studied: modification A and
modification B.
104 plants with modification A were randomly sampled, and 90 plants with modification B were
randomly sampled. These 194 plants were exposed to the same disease under the same
conditions. Whether they were attacked by the disease (susceptible) or whether they remained
unharmed (non‐susceptible) was recorded.
The data for this experiment is contained in dataset_tut_05_gen_mod_plants.sav.
1. Identify the sample(s) and population(s).
There are two samples and two corresponding populations.
One is the sample of size 104, which corresponds to the population of type A modified plants.
The other is the sample of size 90, which is drawn from the population of type B modified plants.
2. Comment on the percentages in the 2 × 2 contingency table of modification type and
susceptibility below.
51.9% of Type A plants are susceptible to the disease, whereas only 33.3% of Type B plants are
susceptible.
The proportion of type A plants that are susceptible to the disease is 1.5 times larger than the
proportion of type B plants. This does seem to suggest that modification B has a stronger
association with resilience to the disease. (we can do a significance test to see if the difference is
significantly large – topic 11)
3. Write down ̂ , the point estimate for the proportion of type A plants that are susceptible to this
disease.
̂ = 0.519 or 51.9%.
1
4. Compute a 95% confidence interval for the true proportion of type A plants that are susceptible
to this disease.
n ̂ (1‐ ̂ = 25.9 ≥ 5. Thus ̂ is approximately normally distributed.
̂ = 54/104 = 0.519,
Thus, the 95% confidence interval for the true proportion of type A plants that are susceptible to
this disease = ̂ margin of error = 0.519 ± 1.96 x sqrt [0.519(1 − 0.519)/104] = (0.421, 0.617)
5. Discuss which of the statements are true or false.
(a) 51.9% of all the type A genetically modified plants are susceptible to the disease. False.
(b) We do not know what proportion of type A plants are susceptible, but we know it is between
42.1% and 61.7%. False.
(c) There is 95% probability that the interval (42.1%, 61.7%) contains the true proportion of type A
plants that are susceptible to disease. False.
(d) We’re 95% confident that the interval (42.1%, 61.7%) contains the true proportion of type A
plants that are susceptible to disease. True.
Please make sure you know the reasons for the above. The tutors would have gone through them
in class.
6. Suppose now that the true proportion p for type A modified plants is in fact 0.519. Then a
randomly sampled type A plant will be susceptible with probability 0.519. 100 samples of size
104 are obtained from type A plants, and they are exposed to the disease once more. The
histogram below shows the frequency of the 100 sample proportions. Comment on the shape of
the histogram.
2
The histogram appears unimodal, symmetric and approximately normally distributed.
It is centered around 0.519.
Most of the observations are between 0.45 and 0.60.
7. 100 confidence intervals for the proportion of type A plants that are susceptible to this disease
(p) have been constructed using each of the 100 samples. Some but not all of the confidence
intervals contain the true value 0.519. Below is the frequency table of the outcome whether the
confidence interval contains 0.519. How many of the intervals contain 0.519? Why is it not exactly
95?
It is not exactly 95% because the confidence level is a long‐term proportion (probability). It does
not mean that exactly 95% of the 100 intervals will contain the true value. It only means that in
the long run, the proportion of intervals that contain the true value will stabilize to 0.95.
8. Suppose we are planning for a new experiment, to test for resistance to a new disease. We wish
to obtain a 95% confidence interval for the true proportion with a total width of less than 0.2.
What is the corresponding maximum allowable margin of error? What is sample size n should we
use?
The total width of a confidence interval is two times of the margin of error, thus the maximum
allowable error is 0.20/2 = 0.10 (10%).
.
The sample size, 96.04 = 97
.
