Description
In the Minnesota barley trials discussed in the multiway chapter of the course notes and in
Cleveland’s book, the major controversy was whether the Morris yields for 1931 and 1932 were
mistakenly reversed. Cleveland asserts that the yields given for Morris 1931 were actually for
Morris 1932, and vice versa. The file minnesota.barley.yield.txt contains a fuller set of information
about the barley trials than is given in Cleveland, with data from 1927 to 1936:
• yield: barley crop yield in bushels per acre (a bushel is four pecks);
• gen: one of a number of types of barley;
• year: a year from 1927 to 1936;
• site: one of six locations in Minnesota.
With more data, we should be able to get a better idea of whether the Morris data can be explained
as random variation or if it’s a mistake.
Questions
1. Draw an appropriate faceted graph showing how barley yield varied by year at each location,
using color as necessary. (Note: Lines might be better than dots here.) When looking at
successive years, was it more common for the yields to move in the same direction at all
locations, or was it more common for the yields to increase at some locations and decrease at
others?
2. Fit a model with yield as the response and gen (variety), year, and site as explanatory
variables, with the goal of determining whether Morris 1931-1932 is an anomaly. Justify why
you chose this model and not some other one. Because of outliers, you should use a robust
fitting method.
3. Draw plots of the fit and/or residuals with the goal of determining whether Morris 1931-1932
is a mistake, or whether it can be explained as natural variation. As best as you can tell, was
there a mistake?
1
