(d) Calculate three new design points along the path of steepest descent.
Problem 4: A forestry researcher is interested in studying the growth of five species of pine
trees. He has 30 experimental plots located throughout the southeastern U.S. Each of these
plots is divided into 5 subplots, in order to create a randomized complete block design within
each of the 30 plots. Let Yijk denote the height of tree species k within subplot j within plot i.
(a) Write down a fixed effects model that contains additive effects for plot, subplot, and
species, as well an an interaction between species and plot. Remember, there are 150
total subplots, not 5!
(b) Write down an additive model that treats the species effect as a fixed effect, and the plot
and subplot effects as random. Make sure you define and specify what all the terms in
the model are.
(c) According to the random effects model from part (b), write down the covariance matrix
for the vector (Y111 , Y112 , Y123 , Y211 , Y311 ). Use your notation from part (b) in answering
this question.
(d) Write down the correlation matrix for the vector from part (c).
Problem 5: We have the following quadratic model:
Yi = 0 + 1 (xi x )2 + i ,
where i = 1, . . . , 11. xi is the covariate, x is the unknown location of the stationary point
(maximum or minimum), and i i.i.d. N(0, 2 ).
(a) Suppose we collect data and conduct and fit the above model, resulting in a residual sum
of squares of 0.335. Then we fit a reduced model supposing that x = 1, resulting in a
residual sum of squares of 0.523. Conduct an F test of whether x = 1 or not. Give the
F statistic and the p-value.
(b) Is 1 in the 95% confidence interval for x ? Why or why not?
(c) Is 1 in the 90% confidence interval for x ? Why or why not?
(d) What is the smallest % confidence interval that 1 belongs to (assuming it is constructed
from an F test as above)?