By the end of this assignment, you should be able to achieve the following tasks in R:

These achievements belong to Learning Outcomes 2, 3, 4, 5, 6.

Click on any blue text to visit the external website.

Note: If you contact me for help or (better yet) open an issue in the public discussion forum, please include the code that is not working and also tell me what you have tried.

Preparation

Wrangling II

Your task here is simple. Write code chunks that

Then, describe the results briefly. Identify patterns and trends, potential outliers, and other results you find interesting. Think like the scientist you are!

You may need to review your Introduction to R coding exercise to meet some requirements, especially factors and how to select one column from a dataframe, and how to change one value to another.

A few things:

Importing Data

As a reminder, The here package automatically sets all search paths relative to the project file. We’re keeping our files to import in a folder called data. Thus, the general format you should use is

my_data <- read_csv(here("data", "my_file.csv"))

here tells read_csv to look for your file in the data folder. This is a simple and portable solution across all computers. Use this format for this and all future assignments.

Source your functions

You can use R’s source() function to run R code contained in an external file. Source your my_functions.R script that you wrote in the previous assignment by including this code chunk. Use the same format you do with your data folder and here() but use your scripts folder instead. You’ll have to change the name of my_functions.R if you did not use that name for your scripts file.

Darters in riffles

Import data

  • Read in darters.csv. These are the same data as an earlier assignment but in csv format.
  • Use filter() to remove tetrazonum and zonale.
  • Do not delete the minsub column like you did in the previous assignment.

Wrangle data

The substrate was measured at each trap site. The relative proportion of the two most commmon substrate types (sand, fine gravel, small gravel, large gravel, and cobble) was estimated. The relative abundance of the most abundant substrate type was recorded as majsub. The less abundant substrate type was recorded as minsub. Together, majsub + minsub = 1. If they do not sum to 1, then there was a data recording error.

  • Use mutate() to perform these wrangling tasks
    • Sum together the majsub and minsub into a new column called total_substrate. Does total_substrate sum to 1 for all observations?
    • Change the riffle values of 1 and 2 to Riffle 1 and Riffle 2.
    • Change the length data from centimeters to millimeters. It is more common to use mm for small fishes.
    • Change the sex values of f and m to Female and Male.
  • Only total_substrate needs a new column. The other changes should use their own columns.

Summarize data

Use group_by() and summarize() to summarize the mean length, depth, and velocity for each species for each riffle. Save this summary to a different object than the original data. You will need both in the following plot.

Table results

  • Use kable to make a table with the summary means.

Graph data

Plot 1

  • Make a plot that uses the stat_summary() technique shown in R4ds section 3.7, reason 3 (scroll down, just before 3.7.1 Exercises) but use x = species and y = length.

    This will plot a point to show the median length, and draw a line from the shortest (min) to the longest (max) length to show the range of lengths for each species.

  • Remember that ggplot2 uses a “layered grammar of graphics.” Add a geom_point() layer to add the mean length of each species to the plot. Use a filled square shape. You will have to use the summarized data set to plot the mean for this geom.

Note: If you do not get a point in the next plot showing the mean for E. blennioides, then you did something wrong. What was it? Find your error and fix it.

  • Turn the plot so that the species are listed on the left axis and length on bottom axis.

  • Add facet_wrap() to separate the two riffles.

  • Add labs() so that length includes the appropriate unit of measurement in parentheses and that the species axis is not labeled (x = NULL).

How does the relationship of the median, mean, and range change between the two riffles? In other words, how does the size distribution change between the two riffles? A correct plot looks this this.

Plot 2

You probably noticed that the location of the median is not centered between the minimum and maximum values, even though the median is the middle value.

You decide that you could display the data more effectively. You want to show the individual measurements and emphasize the mean over the median. You also decide you want to try a slightly different way of graphing the data.

  • Make a new plot with a geom_point() layer of the length variable for each species. Change the point color to a lighter shade in the range of gray50 to gray70. The points should contrast against the overall gray background but they are not the main visual focus of the graph.

  • Add a stat_summary() layer with these arguments:

    • fun = median
    • geom = "point"
    • size = 2.2
    • Do not include fun.min or fun.max arguments.

  • Add another stat_summary() with the same arguments but change the function to mean, size to 3, and add the color = "maroon" argument. If you get a really large point, then you didn’t quite follow the instructions. Read them again carefully and fix it. Your maroon point should be just a bit larger than the other points.

  • Facet, label, and flip the graph as you did above. A correct plot looks like this.

Stress and corticosterones in birds

Beaugeard et al. 2018 (open access) studied the effects of stress from urbaniztion on House Sparrow (Passer domesticus).

Import data

  • load the smatr library when you load tidyverse.
  • Data file: feathers.csv.
  • Tidy the data if and as necessary.
  • The data were recorded by a team of French scientists, so decimal numbers use a decimal comma instead of decimal point (e.g., 2,31 instead of 2.31). Use the locale = locale(<argument>) argument in the read_csv() function to specify the decimal mark. Type ?locale to read the help file. The only argument you need to supply to locale() is for the decimal mark.
  • Their column names include spaces and units of measurement, generally not a good idea because you have to use backticks to reference the columns. Rename the columns when you import them, or rename them after you import the data using the rename() function. Use these names, in order from left to right:
    • capture_date, day, site, urban_score, cortF, cortB, sex, smi, tarsus, mass, basal_cort, stress_cort
  • Hint: If you rename the columns when you read in the data, you will need to use the skip argument to remove the row of column names in the file.

Wrangle data

The authors use an index called the “scaled mass index” (SMI), included as a column. You will use your scaled_mass R function to recreate that value. Recall that the equation to calculate SMI for each individual (\(i\)) is,

\[\mathrm{SMI}_i = M_i \times (L_0/L_i)^b,\]

where \(L_0\) is the mean tarsus length of all birds in the sample, \(M_i\) and \(L_i\) are the mass and tarsus1 length of each individual, and \(b\) is the slope estimate of the standardized major axis regression on the log-transformed mass and log-transformed tarsus length.

Fortunately, you do not need to know the details of the analysis. You have the tarsus lengths and mass from the individual birds in your tibble. You need to calculate \(b,\) the slope of the regression. Follow these steps carefully.

  • Calculate the mean tarsus length of all individuals in the data set, and store it in a variable called mean_tarsus. You should know how to calculate the mean of all values in a column.

  • Use mutate() to log-transform (log()) the body mass (mass) and the tarsus length (tarsus). Save these in new columns as log_mass and log_tarsus, respectively.

  • Run major_axis <- sma(log_mass ~ log_tarsus, data = <data>). Substitute the name of your imported data for <data>. The sma() function calculates the standardized major axis regression.

  • Apply the summary() function to major_axis to see the results of the regression.

    • The value you want for \(b\) is the slope of the estimate.

  • Apply the coef() function to major_axis. The output is a named numeric vector with two elements. The value you want is the second element. Store that value in the variable b. Do not hard code the value. Review Homework 1, Part 2 to recall how to get values from a named vector. Make R do the work for you!

  • Use mutate and your scaled_mass function to calculate SMI and store the values in a new column called new_smi. You need to pass the mass column, the tarsus column, and b to your function.

  • Compare your new column with the column you imported. Your values should be nearly identical, differing by no more than 0.1. Hint (optional): Use the select() function of dplyr to rearrange your columns so that smi and new_smi are side by side. Or, print them out together. Note: If your numbers are not correct, figure out where you made a mistake and correct it.

Summarize data

Next, summarize the basal and stress corticosterone data.

  • Group by site, and then use summarize() to calculate the mean and standard error of the mean \((SE_Y)\) for basal_cort and stress_cort for each site. Use your std_err function to calculate standard error of the mean.

Graph data

Plot 1

  • Make a boxplot of new_smi for each site. Do you see any clear differences among the four sites?

Plot 2 A common graph is to plot the means with error bars based on the standard errors of the means.

  • Make a plot of the basal corticosterone concentration (ng/mg) for each site. Use geom_point(). Add a geom_errorbar() layer. Use the ? function and the interwebs if you need, but here are a few hints:
    • x should be the sites,
    • geom_errorbar() needs ymin and ymax to draw the lower and upper ends of the error bars. Here, ymin should be one standard error below the mean and ymax should be one standard error above the mean. How could you do that? Good. Now do it.

Plot 3

I am not a fan of the Star Wars Tie Fighter look of the default error bars. The horizontal lines do not add any information about the data. In other words, they decrease the data-ink ratio.

Repeat the above plot but with two differences. Use the mean_stress and its standard error, and use geom_linerange() instead of geom_errorbar. Linerange and errorbar use the same arguments.

Graduate and Honors students

You must modify one of the above graphs with the following requirements:

  • use size of 3 for the point located at the mean,
  • color the point maroon,
  • place the error bars behind the points so that a black line does not cut across the face of each point.
  • Assume you like the Tie Fighter look but you do not want them so exaggerated. Size them to something much smaller but still present. The exact size depends on your tastes. Note: I encourage undergrads to attempt this but this does not count for or against your grade.

et Vóila

  1. The tarsus looks like part of the bird’s leg but is homologous to the tarsal bones in the vertebrate foot.↩︎