Lab 7.1 - Faculty Hiring Patterns

This document explores the claim that universities are increasingly hiring more part-time faculty and less full-time faculty. It starts with a rather poorly designed plot, which shows the trend but only at the cost of some considerable effort on the part of the reader, and improves the visualization, which shows the trend more clearly.

Analysing the Original Visualization

The American Association of University Professors (AAUP), a nonprofit membership association of faculty and other academic professionals, created this report. The report discusses trends in instructional staff employees between 1975 and 2011, and contains a visualization very similar to this one:

Instructional staff employment trends

🚧 Analyze this plot: what type of plot is it? What variables are mapped to what aesthetics? What is easy to see, what is hard to see?

Designing Alternate Visualizations

🚧 Sketch on paper two alternative designs for this plot that might do a better job of illustrating the hiring levels different types of faculty over time. Describe the alternatives in the same way as you just did above (type, aesthetic mappings) and explain how they might be improvements of the original.

Rebuilding the Visualization

We now update the visualization incrementally, starting with an approximation of the original.

1. Reproduce the original plot.

🚧 Remake the original plot, starting with this dataset and trying to make it look as follows.

aaup-1

Notes:

• The dataset will need to be pivoted, so that you have columns for faculty_type, year, and percentage.
• Use a column chart with dodged bars (geom_col(position = "dodge")).
• Use the Viridis color scale (scale_fill_viridis_d())
• Include labels (labs()) for: x, y, fill, title, caption

2. Convert to a stacked bar plot.

We’d like to more directly compare the hiring levels of the different faculty types.

🚧 Update the plot to look like this.

aaup-2

Notes:

• See if you can do this without repeating the pivot_longer.
• We used theme_minimal().

3. Convert to a line plot.

Line plots tend to be good for view values over time.

🚧 Update the plot to look like this.

aaup-3

Notes:

• Note the use of a redundant encoding (shape). I had to add a geom_point layer to draw those.
• Note that, at first, the lines don’t actually show up. You can hack you way to make it work by adding group = year to the aesthetic mappings; a Google search on the error message you get will probably get you there. But that yields the plot shown above. If you look carefully, you’ll see something badly wrong with it.
• Once you notice what’s wrong with the line plot, then fix things by making sure that the year is treated as.numeric. (And once you do that, you don’t need the group = year hack anymore anyway.)
• If you made changes to earlier code, make sure your earlier plots still work. (You may need to replace year with factor(year) in those plots, depending on how you did it.)

4. Just show the numbers.

Sometimes, simply displaying numbers is as effective as visualizing them. Let’s make the table below, which shows the changes in hiring levels, i.e., the 2011 percentage minus the 1975 percentage.

🚧

• First, make the table the easy way, starting with staff. Tip: 2011 isn’t normally a valid variable name, but you can use backticks (next to the 1 key): `2011`. You’ll probably need to select some columns.
• Once you have the resulting table, can you now make the same thing but starting with your long-format data?
  • Yes, you’ll need to pivot.
  • Can you do write this pipeline without using select()? Hint: I think that select and filter should have been called select_columns and select_rows.

faculty_type	1975	2011	change
Full-Time Tenured Faculty	29.0	16.7	-12.3
Full-Time Tenure-Track Faculty	16.1	7.4	-8.7
Full-Time Non-Tenure-Track Faculty	10.3	15.4	5.1
Part-Time Faculty	24.0	41.3	17.3
Graduate Student Employees	20.5	19.3	-1.2

Notes:

• You can format the table by piping the resulting dataframe through knitr::kable().

Drawing Conclusions

🚧 In conclusion, are universities hiring more part-time faculty and less full-time faculty? Why might this be?