Worksheet 8

Published

February 21, 2025

Packages

library(tidyverse)
library(car)
library(lme4)

One not especially long scenario, but you may find yourself with some thinking to do.

You have intelligent rats

Each of 12 students trained rats to run a maze, and the number of successful runs (out of 50 attempts) was recorded on each of 5 days. Some of the students were told that the rats they were training were “bright” (intelligent) and some of them were told that their rats were “dull” (not intelligent). In actual fact, all the rats were from the same source, and none of them were any more intelligent than the others. Did it make a difference whether the students were told that their rats were intelligent on the number of mazes the rats successfully ran, and, if so, was the effect consistent over time? The same group of rats were trained by the same student throughout this experiment, so it makes sense to treat the data as repeated measures.

The data are in http://ritsokiguess.site/datafiles/intelligent-rats.csv. The columns of interest to us are:

  • Student: a numerical identifier for each student
  • Treatment: what the student was told about their rats
  • Day1 through Day5: the number of successful runs on each day.

There are some other variables that will not concern us here.

  1. Read in and display (some of) the data.
  1. Set up and run an appropriate repeated-measures ANOVA. What do you conclude from it? (Hint: set up the response variable, and go through the steps required to run the Manova. Obtain the appropriate P-values, describing your process.)
  1. Draw a suitable interaction plot (for treatment and time). How does this clarify your conclusions of the previous part? (Hint: you’ll need longer data. If you want to be consistent with me, use Day for the names of the days, and runs for the numbers of runs.)
  1. For each time point (day), run a simple effects analysis: that is, using an ordinary aov, test whether the number of successful runs depends on whether the rat was labelled as “dull” or “bright”.
  1. Do a mixed model analysis (hint: use the long data that you made for your interaction plot.)
  1. Compare your findings from the mixed model to the ones from the Manova analysis you did earlier.