Biostatistics for Pre-Health: Week 5 in-class exercises
Please answer the questions as thoroughly as you can. Some questions have definite
answers, but others are more subjective. I want you to think about the issues and
try to formulate conclusions. Please type your answers in a separate (Word or similar)
document, using complete sentences.
For these exercises, please work in groups of 3-4 students. Please turn in written
solutions, in a Word file, as a group, using the Moodle submission.
For all of the following questions, refer to the article of Bangalore et al. First read
the article (it’s quite short); then answer the following questions.
1. What are the research questions addressed by the study?
2. What is the sample size? What measures are collected for each member of the
sample? Be careful! Not all studies of human behavior have to be carried out
at the person-level. Studies carried out at higher levels of organization are called
ecological studies.
3. What statistical methods are used in the article?
Now refer to the Excel file [Link], which contains the study data.
All charts produced below should have informative axis labels and titles.
4. Create a new variable log(FR Deaths) that is equal to the natural log of the
variable Firearm-Related Deaths. This is exactly the same as what has al-
ready been done for Guns per 100. Now make informative histograms of both
Firearm-Related Deaths and log(F-R Deaths). Include both charts in your
solutions. Can you explain why the log-transformed variable might be better for
analysis?
5. Make a scatterplot of log(Guns) (on the x-axis) against log(F-R Deaths) (on
the y-axis), and compute the correlation r between these two variables. Include
the chart in your solutions. Copy and paste in a complete sentence from the
paper that confirms your findings.
6. Run a multilinear regression to predict log(F-R Deaths) from log(Guns) and
Mental illness per 100,000, and include the relevant output. Make sure that
you ask for and include residual plots.
(a) What is the best-fit equation?
(b) What are the p-values for the two predictors?
(c) What are the R2 and Adjusted R2 for the analysis?
(d) Confirm your results from the last two parts with those presented in the
paper. Do they agree?
1
(e) Write out your conclusions in sentence format. Include coefficients, with
95% confidence intervals, in your summary.
(f) Explain what the residual plots show. Which countries look like outliers in
this analysis?
7. Calculate the correlation r between Mental illness and log(F-R Deaths), and
confirm with the article Figure. Also make a scatterplot of Mental Illness
against log(F-R Deaths), including a best-fit line, using the supplied data.
How does you scatterplot compare with the one in the article (Part C of the
Figure)? Is the existence of outliers as clear?
8. Formulate one more hypothesis that could be addressed with this data set. Make
appropriate charts and compute relevant descriptive and inferential statistics.
Write up your conclusions in sentence format!
For this week’s Software test, complete the following and submit it individually.
1. Load the [Link] file. This file is already in R format and includes
the log(F-R Deaths) variable.
2. Compute the correlation between log(Guns) and log(F-R Deaths). Include
the output in your solutions.
3. Run a multilinear regression to predict log(F-R Deaths) from log(Guns) and
Mental illness per 100,000. Use the example R script as a template.
• Ask for diagnostic plots.
• Find confidence intervals for the coefficients.
Include all of the output in your solutions. But you do not need to summarize
in words. (After all, these should be the same results as obtained in Excel, and
you already did that!)