Homework #9 :

Scientific Statistics : Math 1400

Problems

1. Visualize this! In pre-covid times, I mentioned that we are cognitively bound to locate all phenomena
in time, space, and function. This can also be a helpful way of analyzing visualizations and their
effectiveness. A good figure should arrange these features of phenomenon in such a way that the
viewer can tell as story about what is happening.
• Examine the visualizations found in the folder linked here: link. Remember what is space, what
is function, and what is time maybe creatively used (or absent) for a given plot.
• For each, state how time, space, and function are organized. Given a one sentence summary of
the story for each.
• Recall your scientific measurement that you were exploring for your writing project. At the core
of the project is a measurement: some scientific assay that defines how we currently think about
how to measure that quantity of interest. Create a figure of this key process of measurement,
keeping in mind how time, space, and function are organized. Write a one paragraph summary
of how the measurement works.
2. Probably an error? Suppose you conduct an experiment N = 90 times and find that the mean is 5.5.
The observational variance is known to be 2.5. You have good reason to believe that the underlying
distribution of the experiment is normal with mean 5.
• What is the test statistic? What is the sampling distribution?
• What is the null hypothesis? What is a reasonable alternative hypothesis?
• What is the null distribution?
• What is the probability that you observe your data or something more extreme (further from the
center of the distribution) given your assumptions?
• Draw a picture of the null distribution, the test statistic, and p-value on this plot. Suppose you
set α = 0.05, where would the critical region be?
3. A simple hypothesis, repeated. On Blackboard is a data set multiple.csv. There is a header
and the file is comma-separated. There are 30 experiments, each with 50 replicates.
• Read about the boxplot here: link.
• Make a set of boxplots with data using the boxplot function and add a color to them of your
choosing. Add a horizontal line at 0. Do any of the experiments appear to have means different
from zero? (Just by eye.)
• For each column in the data set, calculate the z-score with the assumption that µ = 0.
• For which groups, would you conclude µ 6= 0 if using α = 0.05 as a significance level? Plot the
z-scores and color all those for which you would reject the null hypothesis in red.
• Would you believe any of these results? All of these results? Why or why not?
4. Two for two? You sample X1 , · · · , X100 i.i.d from population A and find a mean of 10. You sample
Y1 , · · · , Y64 i.i.d from population B and find a mean of 11. The standard deviation is the same for each
population and is 2.

1
 • What is the distribution of X A − Y B ?
• What is the test statistic and the corresponding sampling distribution?
• Suppose the null hypothesis is that µA = µB . What is the null distribution?
• Use R to calculate the p-value associated with the observed value. Would you reject at the α = 0.01
level?
5. Effect of effect size? Suppose you collect N = 25 samples from a population that is normally
distributed with mean µ and σ 2 = 25. You perform a hypothesis test assuming that µ is 0, when in
fact it is 1. You use α = 0.01 as your significance level. The effect size is the difference between the
true and the hypothesized mean.
• Draw a picture of the null distribution and the true sampling distribution (that is, the sampling
distribution for the actual value of µ). Use qnorm to find the critical value. Mark the critical
region of null distribution.
• What is the probability that a draw from the true sampling distribution will fall into the critical
region? What does this tell you about the probability of rejecting the null hypothsis?
• Re-do the first two bullets with N = 100.
6. Binomial rejections? As we started in our last assignment, we can to use the power of R to examine
how probable it is that we’ll reject the null hypothesis when the null hypothesis is either true or false
when looking at the binomial distribution. We will assume that the number of coin tosses is N = 20,
that the random variable X takes a value k, α = 0.05, and that the null hypothesis is p = 0.5 and the
alternate hypothesis is that HA : p > 0.5
• Using qbinom, find what value of k you would need to observe to reject the null hypothesis given
that p = 0.5. Call this value kc for critical k.
• Make a plot of the null distribution, with the values in the critical region colored in red, with the
rest in black.
• Now use pbinom to work out the probability that you observe kc or something greater when
p = 0.55 and p = 0.75. How does this relate to the probability of rejecting H0 : p = 0.5 if the
true value of p is 0.55 and 0.75?