Upload
vutruc
View
241
Download
8
Embed Size (px)
Citation preview
Graphical Gems in the agridat PackageKevin Wright2017-11-29
Abstract
The agridat package is an extensive collection of data sets that have been previously published in booksand journals, primarily from agricultural experiments.
This vignette presents graphical views of a few of the datasets in this package.
Setup
This exhibit of agricultural data uses the following packages.
library("agridat")library("desplot")library("gge")library("HH")library("lattice")library("latticeExtra")library("mapproj")library("maps")library("reshape2")
1
Potato blight incidence over space and time
lee.potatoblight
1983−10−17 1983−10−25 1983−10−31 1983−11−07 1983−11−14 1983−11−21 1983−11−28 1983−12−05 1983−12−12 1983−12−20
1985−11−18 1985−11−25 1985−12−02 1985−12−09 1985−12−16 1985−12−23 1985−12−30
1987−11−16 1987−11−23 1987−11−30 1987−12−07 1987−12−14 1987−12−21 1987−12−28 1988−01−05
1991−11−08 1991−11−15 1991−11−27 1991−12−05 1991−12−24 1992−01−07
1993−11−19 1993−11−29 1993−12−10 1993−12−22 1993−12−31
1995−11−24 1995−12−01 1995−12−08 1995−12−14 1995−12−22
1997−11−04 1997−11−19 1997−11−29 1997−12−05 1997−12−12 1997−12−22
1999−12−07 1999−12−16 1999−12−30 2000−01−07 2000−01−15
2001−11−13 2001−11−23 2001−11−29 2001−12−06 2001−12−18
2003−12−08 2003−12−15 2003−12−23 2003−12−31 2004−01−16
2005−11−07 2005−11−15 2005−11−29 2005−12−15 2005−12−23
0
2
4
6
8
Lee (2009) analyzed a large dataset to evaluate the resistance of potato varieties to blight. This data containsevaluations of a changing set of varieties every two years, evaluated in 5 blocks, repeatedly throughout thegrowing season to track the progress of the disease. Each panel shows a field map on the given date, with aseparate row of panels for each year.
Would you include field spatial trends in a model for these data?
2
lee.potatoblight 1983
Date
Blig
ht r
esis
tanc
e sc
ore
2468 1 1
11
11
1 1 1 1
2 22
22
2 2 2 2 2
3 33
33
3 3 3 3 3
4 44
44
4 4 4 4 4
55 5
5
5 5 55 5 5
060.11 1 1 1 1 1 1 1
1 12 2 2 2 2 2 2 2 2
23 3 3 3 3 3 3
33
3
4 4 4 4 4 4 44
44
5 5 5 5 5 5 55 5
5
064.11 1 1 1 1 1
11
11
2 2 2 2 2 22 2
22
3 3 3 3 3 33
33
3
4 4 4 4 4 44
44
4
5 5 5 5 5 55 5
5
5
064.181 1
11
11 1 1 1
1
2 22
2
2 22
2 2 2
33
33
33
3 3 3 3
4 44
4
44 4 4 4 4
5 55
5
55
5 5 5 5
064.211 1 1 1 1 1 1 1
11
2 2 2 2 2 22 2
2
2
3 3 3 3 33
33
33
4 4 4 44
44 4
4
4
5 5 5 5 5 5 55
5
5
064.24
2468 1 1 1 1 1 1 1 1
1 12 2 2 2 2 2 2 2 2
23 3 3 3 3
3 33
3
3
4 4 4 4 44 4 4
44
5 5 5 5 5 5 5 5 55
064.371 1 1 1
1 11
11
1
2 2 22 2
2
2 22
2
3 3 3 33 3
33
33
4 4 4 44 4
44
44
5 5 5 55
55
55
5
065.271
1 11
11
1 11 1
2 22
22
22 2 2 2
3 33
33
33 3 3 3
4 44
4
44 4 4
4 4
5 55
5
55
5 5 5 5
1015.471 1
1
11
11 1 1 1
2 22
22
22 2 2 2
3 33
33 3 3
33
3
4 44
4
4 4 4 4 4 4
5 55
55 5 5 5 5 5
1053.571 1 1 1 1 1 1 1 1
12 2 2 2 2 2 2
2 22
3 3 3 3 3 3 33
33
4 4 4 4 4 4 44
44
5 5 5 5 5 5 55
55
1060.9
2468 1 1
11 1
11
1 11
2 22
22 2
22 2
2
3 33
33 3 3
3 3 3
44 4
4
44 4 4 4
4
5 55
55
55 5 5
5
1067.161 1
11
11
1 1 1 1
2 22
22
2 2 2 2 2
33
33
33 3 3 3
3
4 44
4
44 4
4 4 4
5 55
5
5 55 5
5 5
826.41 1
11
11
11
1 1
2 22
22
22 2
22
3 33
33
3 33
33
4 44 4
4 4 4 44
4
5 55
55
5 5 55
5
887.891 1
1
11 1 1 1 1 1
2 22
22 2 2 2 2 2
3 33
3
3 3 3 3 3 3
44
4
44 4 4 4 4 4
5 55
5
5 5 5 5 5 5
978.31 1
11
11 1
1 1 1
2 22
2
22
2 2 2 2
3 33
33 3
3 3 3 3
4 44
4
44 4 4 4 4
5 55
5
55 5
5 5 5
993.60
2468
Oct
15
Nov
01
Nov
15
Dec
01
Dec
15
1 11
11
1 1 1 1 1
2 22
2
22 2 2 2 2
3 33
33 3 3 3 3 3
44
4
44 4 4 4 4 4
5 55
5
5 5 5 5 5 5
I.HARDY
Oct
15
Nov
01
Nov
15
Dec
01
Dec
15
1 11
11
1 11
1 1
2 22
22 2
2 22 2
33 3
33
33 3
3 3
4 4 44
44
4 44 4
5 55
55
55
5 5 5
IWAO
ct 1
5
Nov
01
Nov
15
Dec
01
Dec
15
1 1 11 1
1 11
11
2 22
22
2 22
2 2
3 3 33
33
3 3 33
4 44
44 4
4 4 44
5 5 55
55 5 5
55
RUA
Oct
15
Nov
01
Nov
15
Dec
01
Dec
15
1 11
1 11
1 1 11
2 22
22
2 2 22
2
3 33
33
3 3 3 33
4 44
44 4 4 4
44
5 55
55 5 5 5
55
TEKAU
Oct
15
Nov
01
Nov
15
Dec
01
Dec
15
1 1 11
11
1 11
1
2 2 2 22
2 22
22
3 3 3 33 3
33
3
3
4 4 4 44
4 44
44
5 5 5 55
5 55
55
WHA
In 1983, 20 varieties were evaluated in 5 blocks (shown by colored numbers) throughout the growing seasonfor disease resistance. Resistance scores start at 9 for all varieties (shown in panels). As the growing seasonprogresses, the ‘I.HARDY’ variety succumbs quickly to blight, while ‘IWA’ succumbs steadily, and ‘064.1’resists blight until near the end of the season.
Does this view show differences between blocks?
3
An informative prior
harrison.priors
Daidzein level
Berman−ABerman−U
DeviHutabaratLundry−ALundry−UMorrisonPrimomo
SeguinZhou−1Zhou−2Zhou−3Zhou−4Zhou−5Zhou−6Zhou−7Zhou−8Zhou−9
priorConstructed
500 1000 1500
Harrison, Culp, and Harrigan (2012) used a Bayesian approach to model daidzein levels in soybean samples.From 18 previous publications, they extracted the published minimum and maximum daidzein levels, and thenumber of samples tested. Each line in the dotplot shows large, dark dots for one published minimum andmaximum. The small dots are imputed using a lognormal distribution.
All observed/imputed data were then used to fit a common lognormal distribution that can be used as aninformative prior. The common prior is shown by the density at the top of the dotplot.
Do you think it is better to use a non-informative prior, or this informative prior?
4
Data densities for a binomial GLM
mead.germination
Log concentration
See
ds g
erm
inat
ing
(out
of 5
0).
Bin
omia
l den
sity
.
0
10
20
30
40
50T1
−4 −2 0 2 4
T2
−4 −2 0 2 4
T3
0
10
20
30
40
50T4
Mead, Curnow, and Hasted (2002) present data for germination of seeds under four temperatures (T1-T4)and four chemical concentrations. For each of the 4*4=16 treatments, 50 seeds were tested in each of fourreps. In the graphic, each point is one rep. The blue line is a fitted curve from a GLM with Temperature asa factor and log concentration as a covariate. The gray lines show the central 95 percent of the binomialdensity at that position.
Does this display help you understand the logit link and changing shape of the binomial density?
5
Verification of experiment layout
gomez.stripsplitplot
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P1
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
P2
rep
genG1G2G3G4G5G6
nitro0
60120
plantingP1 P1P2 P2
K. Gomez and Gomez (1984) provide data for an experiment with 3 reps, 6 genotypes, 3 levels of nitrogenand 2 planting dates. The experiment layout is putatively a ‘’split strip-plot”. To verify the design, thedesplot package is used for plotting the design of field experiments.
How is the design different from a ‘’split-split-plot” design?
6
Visualizing main effects, two-way interactions
gomez.splitsplit
R1
R2
R3
yield ~ rep | rep
0 50 80 110
140
yield ~ nitrogen | rep
Inte
nsiv
eM
inim
umO
ptim
um
yield ~ management | rep
V1
V2
V3
46810
yieldyield
yield ~ gen | rep
yield ~ rep | nitrogenyield ~ nitrogen | nitrogenyield ~ management | nitrogen
46810
yieldyield
yield ~ gen | nitrogen
yield ~ rep | managementyield ~ nitrogen | managementyield ~ management | management
46810
yieldyield
yield ~ gen | management
yield ~ rep | genyield ~ nitrogen | genyield ~ management | gen
46810
yieldyield
yield ~ gen | gengen
V1V2V3
managementIntensiveMinimumOptimum
nitrogen05080110140
repR1R2R3
Heiberger and Holland (2004) provide an interesting way to use lattice graphics to visualize the main effects(using boxplots) and interactions (using interaction plots) in data. Rice yield is plotted versus replication,nitrogen, management type, and genotype variety. Box plots show minor differences between reps, increaingyield due to nitrogen, high yield from intensive management, and large differences between varieties.
Do you think interaction plots show interaction (lack of parallelism)?
7
3D yield response to fertilizers
Sinclair et al. (1994) examined clover yields as a function of sulfur and phosphorous fertilizer in a factorial-treatment experiment. Dodds, Sinclair, and Morrison (1996) modeled the yield response using a Mitzerlisch-likeequation that allows interacting curvature in two dimensions x and y:
yield = α ∗(
1 + β ∗(σ + τ ∗ xx+ 1
)y)∗
(1 + δ ∗
(θ + ρ ∗ yy + 1
)x)The blue dots are observed data, and the tan surface is the fitted surface drawn by the rgl package).
How would you decide the optimal fertilizer levels?
8
Mosaic plot of potato damage from harvesting
keen.potatodamage
Energy / Genotype
Rod
/ W
eigh
t
E1 E2
R1
R2
R3
R4
R5
R6
R7
R8
G1 G2 G3 G4 G5 G6
W1
W2
W3
W1
W2
W3
W1
W2
W3
W1
W2
W3
W1
W2
W3
W1
W2
W3
W1
W2
W3
W1
W2
W3
G1 G2 G3 G4 G5 G6
Keen and Engel (1997) looked at damage to potatoes caused by lifting rods during harvest. In this experiment,eight types of lifting rods were compared. Two energy levels, six genotypes and three weight classes wereused. For each combinations of treatments, about 20 potato tubers were rated as undamaged (D1, yellow) toseverely damaged (D4, red). Counts per treatment are shown in a mosaic plot.
Which style of lifting rods cause the least/most damage to potatoes?
9
Yield vs covariate for lattice::barley
minnesota.barley
Weather covariate
Bar
ley
yiel
d
103050
200 400 600 8001000
7 9 01
23
45
6Cro
okst
on
1000 1500 2000
7901
2 3
45
6
5 10 15 20 25
7901
23
45
6
7
89 0123
4
56
Dul
uth 7
8901 2
3
4
56 10
30507
890 12
3
4
56
103050
78 90
123
45
6
Gra
ndR
apid
s
78 90
123
45
67
89 01 2
3
45
6
78 9 01
2 56M
orris 7
890 12 5
61030507
89 012 5
6
103050 7
8
90 12
3
4
5
6StP
aul 7
8
90 12
3
4
5
6
78
901 2
3
4
5
6
78
9 01
23
4
56
cddW
asec
a
7890
12
3
4
56
hdd
1030507
89 0
12
3
4
56
precip
Wright (2013) investigated the lattice::barley data. The original two years of data were extended to 10years (from original source documents), and supplemented with weather covariates for the 6 locations and 10years. Each panel shows a scatterplot and regression for average location yield verses the weather covariate.Horizontal strips are for locations, vertical strips are for covariates: cdd = Cooling Degree Days, hdd =Heating Degree Days, precip = Precipitation). Higher values of heating imply cooler weather. Each plottingsymbol is the last digit of the year (1927-1936) for that location.
Does barley yield better in cooler or warmer weather?
10
GGE biplot
PC 1 (32% TSS)
PC
2 (
12%
TS
S)
crossa.wheat
method=svd, center=TRUE, scale=FALSE, missing: 0%
AKAS
BJ
CAEB
EG
ES
ID
ILJM
KN
MGMM
MS
NBPA
PI
SCSE
SG
SJSRSS
TB
TC
Grp1
Grp2
G01
G02
G03G04
G05
G06
G07
G08
G09
G10
G11
G12G13
G14
G15
G16
G17
G18
Laffont, Wright, and Hanafi (2013) developed a variation of the GGE (genotype plus genotype-by-environment)biplot to include auxiliary information about a block/group of environments. Each location is classified intoone of two mega-environments (colored). The mosaic plots partition variation simultaneously by principalcomponent axis and source (genotype, genotype-by-block, residual).
Which genotypes are best to each mega-environment?
11
Nebraska farming income choropleth
nebraska.farmincome
Inco
me
($10
00)
per
coun
ty
crop
animal
100
200
300
400
500
600
700
The Red-Blue palette in the RColorBrewer package is a divergent palette with light colors near the middleof the scale. This can cause problems when there are missing values, which appear as white (technically, thebackground). In order to increase the visibility of missing values, the agridat package uses a Red-Gray-Bluepalette, with a gray color that is dark enough to clearly distinguish missing values.
How does the outlier county (Butler) in northeast Nebraska limit interpration of spatial patterns in the data?
12
nebraska.farmincome
Inco
me
($10
00)
per
squa
re m
ile (
perc
entil
e br
eaks
)
crop.rate
animal.rate
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Because counties are different sizes, the second graphic uses an income rate per square mile. Because of theoutlier, it might be smart to use percentile break points, but doing so hides the outlier. Instead, the breakpoints are calulated using a method called Fisher-Jenks. These break points show both the outlier and thespatial patterns. It is now easy to see that northwest (Sandhills) Nebraska has low farming income, especiallyfor crops. Counties with missing data are white, which is easily distinguished from gray.
Where are farm incomes highest? Why?
13
Las Rosas yield monitor
lasrosas.corn grain yield (qu/ha)
Longitude
Latit
ude
−33.052
−33.051
−33.050
−33.049
−63.848 −63.846 −63.844 −63.842
1999
−63.848 −63.846 −63.844 −63.842
2001
20
40
60
80
100
120
lasrosas.corn experiment design
−33.0520
−33.0515
−33.0510
−33.0505
−33.0500
−33.0495
−33.0490
−63.848 −63.846 −63.844 −63.842
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++−−−−−−−−−−−−−−−−//////////// ///////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++−−−−−−−−−−−−−−−−−−/////////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++−−−−−−−−−−−−−−−−−−/////////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++−−−−−−−−−−−−−−−−−−−////////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++−−−−−−−−−−−−−−−−−−−////////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++−−−−−−−−−−−−−−−−−−−////////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++−−−−−−−−−−−−−−−−−−−−///////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++−−−−−−−−−−−−−−−−−−−−///////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−//////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−//////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−/////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−/////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−−////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−−− ////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−−−///////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−−−///////////////////////
++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−−−////// ///////
−/////// ////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++
// /////// /
1999
−63.848 −63.846 −63.844 −63.842
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++−−−−−−−−−−−−−−−−−−///////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++−−−−−−−−−−−−−−−−−−///////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++−−−−−−−−−−−−−−−−−−//////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−/////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−/////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−/////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−/////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−/////////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−///////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−///////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−///////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−///////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\\+++++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−//////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−−/////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−−/////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−−/////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−−/////////////////////
\\\\\\\\\\\\\\\\\\\\\\\\\++++++++++++++++++++++++++++−−−−−−−−−−−−−−−−−−−−−−/////////////////////
2001
Anselin, Bongiovanni, and Lowenberg-DeBoer (2004) and Lambert, Lowenberg-Deboer, and Bongiovanni(2004) looked at yield monitor data collected from a corn field in Argentina in 1999 and 2001, to see how yieldwas affected by field topography and nitrogen fertilizer. The figures here show heatmaps for the yield eachyear, and also the experiment design (colors are reps, shades of color are nitrogen level, plotting character istopography).
14
Which year showed greater spatial variation in yield?
Time series of corn yields by state
nass.corn
Year
Mill
ion
acre
s of
cor
n
0
5
10
Alabama
1900 2000
Arkansas Colorado
1900 2000
Georgia Illinois
1900 2000
Indiana
Iowa Kansas Kentucky Louisiana Maryland
0
5
10
Michigan
0
5
10
Minnesota Mississippi Missouri Nebraska New York North Carolina
North Dakota Ohio Oklahoma Pennsylvania South Carolina
0
5
10
South Dakota
0
5
10
1900 2000
Tennessee Texas
1900 2000
Virginia Wisconsin
15
The National Agricultural Statistics Service tracks the total number of acres planted to corn (and other crops)for each state in the U.S. There are large changes over the past century in corn acreage for selected states.
Which states were in the corn belt in 1925?
Which states were in the corn belt in 2000?
References
Anselin, Luc, Rodolfo Bongiovanni, and Jess Lowenberg-DeBoer. 2004. “A Spatial Econometric Approach tothe Economics of Site-Specific Nitrogen Management in Corn Production.” American Journal of AgriculturalEconomics 86 (3): 675–87. doi:10.1111/j.0002-9092.2004.00610.x.
Dodds, KG, AG Sinclair, and JD Morrison. 1996. “A Bivariate Response Surface for Growth Data.” FertilizerResearch 45 (2): 117–22. doi:10.1007/BF00790661.
Gomez, K.A., and A.A. Gomez. 1984. Statistical Procedures for Agricultural Research. Wiley-Interscience.
Harrison, J.M., D. Culp, and G.G. Harrigan. 2012. “Bayesian MCMC Analyses for Regulatory Assessments ofFood Composition.” In Kansas State University Conference on Applied Statistics in Agriculture, Manhattan,Kansas.
Heiberger, Richard M, and Burt Holland. 2004. Statistical Analysis and Data Display: An IntermediateCourse with Examples in S-Plus, R, and SAS. Springer.
Keen, A., and B. Engel. 1997. “Analysis of a Mixed Model for Ordinal Data by Iterative Re-WeightedREML.” Statistica Neerlandica 51 (2): 129–44. doi:10.1111/1467-9574.00044.
Laffont, Jean-Louis, Kevin Wright, and Mohamed Hanafi. 2013. “Genotype Plus Genotype× Block ofEnvironments Biplots.” Crop Science 53 (6): 2332–41. doi:10.2135/cropsci2013.03.0178.
Lambert, Dayton M, James Lowenberg-Deboer, and Rodolfo Bongiovanni. 2004. “A Comparison of FourSpatial Regression Models for Yield Monitor Data: A Case Study from Argentina.” Precision Agriculture 5(6): 579–600. doi:10.1007/s11119-004-6344-3.
Lee, Arier Chi-Lun. 2009. “Random Effects Models for Ordinal Data.” PhD thesis, University of Auckland.
Mead, R., R.N. Curnow, and A.M. Hasted. 2002. Statistical Methods in Agriculture and Experimental Biology.9th ed. CRC Press.
Sinclair, AG, WH Risk, LC Smith, JD Morrison, and KG Dodds. 1994. “Sulphur and Phosphorus in BalancedPasture Nutrition.” In Proceedings of the New Zealand Grassland Association, 56:13–16.
Wright, Kevin. 2013. “Revisiting Immer’s Barley Data.” The American Statistician 67: 129–33.doi:10.1080/00031305.2013.801783.
16