Overview

In your second mini-project, you’ll apply what you’ve learned about communication to write a walk-through of fitting a model to data. This can be “blog-post” style, “journal article” style, or any other style you like!

  1. Choose an audience for your project!

    • Stat audience: writing for someone who is trying to learn more about the model that you’re fitting, and is maybe less interested in the code or the specific dataset

    • Coding audience: writing for someone who is trying to learn how to fit the model and make the graphs that you’re using, and is maybe less interested in the statistical details or the dataset

    • Subject area audience: writing for someone who is interested in your dataset and conclusions, rather than the nuts and bolts of what you did. You should still include the models, code, and graphs, but will likely take less time and space explaining them.

  2. You’ll write up an overview of an analysis for the audience you’ve chosen. This should have code and graphs interspersed throughout the document.

Options and Examples

You have a few different options for this project, ranging from writing up one of the labs from the week, to doing an entirely “from scratch” analysis on a dataset of your choosing:

  • Expand one of the labs from this week (05-08).

    • Use the COVIDCAST data from Labs05 and 08, but choose different signals (run unique(covid$signal) to see options) and include at least two predictors in your model
    • Lab06: Fit a model to variables in the gapminder data
    • Lab07: TBD

  • MDSR Ch9 Problem 7

  • MDSR Ch10 Problem 6

  • Fit a t-test to Blood Pressure Data

    • Called nhanes_2015-2016 on moodle.

    • Test whether systolic blood pressure differs between males and females.

    • Check the data documentation file for more info

  • Run an ANOVA on the Technical Difficulties Data

    • Called gss2018 on moodle

    • GSS Data explorer

    • Test whether the mean time spent on technology use (USETECH) is equal across degree groups (DEGREE)

    • Check the data documentation file for more info

  • Fit a Linear Regression model to the Needle Exchange Data

    • Called dist_ssp_amfar on moodle

    • See how distance to nearest syringe program (dist_SSP) is related to percent uninsured (pctunins)

    • Check the data documentation file (opioid county codebook) for more info

  • Fit a Logistic Regression model to the Libraries Data

    • Called pew_libraries_2016_cleaned on moodle

    • See how library use (uses.lib) is related to predictors like age and sex

    • Check the data documentation file for more info

  • Your final project dataset

  • Any other dataset

I expect most projects to do some type of regression, but you are also welcome to perform a more basic statistical analysis, like a t-test or ANOVA, if it is appropriate for your data!

Submission

Your submission should include:

  • Your .rmd file including all analyses and text
  • Your final knitted report (in the format of your choosing)
  • Any necessary data files

Your report should include:

  1. An introduction
  2. Explanations of each step of your analysis
  3. All necessary graphics
  4. A short conclusion

(This is intentionally vague, since outlines will vary widely depending on the type of analysis you choose.)

All of your graphs should be integrated within your text, and you should pay attention to the formatting of the entire document.

Rubric

A satisfactory project will:

  • Include all necessary code for recreation
  • Be well-written, pitched towards the chosen audience, and roughly 1000 (~2 single spaced pages) words in length
  • Meet minimum submission quality for graphics (no typos, appropriate labels and font sizes, a readable theme)

An excellent project will meet all of the requirements for a satisfactory project, plus:

  • Choose a dataset from outside of the labs OR
  • Use a modeling framework not discussed in class or used in labs (some examples: k-means clustering, hierarchical clustering, weighted regression, Poisson regression, LASSO regression)

AND

  • Be publication-ready/especially polished.

Please include a self-grade (with explanation) at the end of your report.