Learn R Programming

agricolae (version 1.3-8)

design.bib: Randomized Balanced Incomplete Block Designs. BIB

Description

Randomized balanced incomplete block designs are an important tool in agricultural experimentation, especially when experimental conditions prevent the inclusion of all treatments within each block. Balance can be "exact" or "approximate"; the requirement for exact balance can significantly increase the number of blocks and experimental units needed. The randomization option ("Random") employs Rs number-generation methods, with the seed set using the set.seed(seed, kinds) function.

Usage

design.bib(trt, k, r.max=NULL, serie = 2, seed = 0, kinds = "Super-Duper",
balance = c("exact", "approximate"), iterations = 3000L)

Value

parameters

Design parameters

statistics

Design statistics

sketch

Design sketch

book

Fieldbook

concurrences

Matrix containing the count of pairs; the diagonal shows the repetitions

pair.frequency

Frequency of pairs of treatments compared in each block

replication

The replicates calculated for each treatment

Arguments

trt

Treatments

k

size block

r.max

repetition maximum

serie

number plot, 1: 11,12; 2: 101,102; 3: 1001,1002

seed

seed

kinds

method for to randomize

balance

exact or approximate

iterations

3000L

Author

Felipe de Mendiburu

Details

if r.max = NULL, then it calculates the value of r smaller for k defined. In the case of r.max = value, then the possible values for "r" is calculated

K is the smallest integer number of treatments and both values are consistent in design.

kinds <- c("Wichmann-Hill", "Marsaglia-Multicarry", "Super-Duper", "Mersenne-Twister", "Knuth-TAOCP", "user-supplied", "Knuth-TAOCP-2002", "default" )

balance <- c("exact", "approximate")

iterations <- 3000L. Process control for the approximate completion

Methodological Rationale for Extending the design.bib() Function

These designs are characterized by an equal number of replications per treatment and a constant concurrence frequency for each pair of treatments (Bailey, 2008).

However, the requirement for exact balance can substantially increase the number of blocks and experimental units needed, limiting their applicability in situations where resources are constrained.

The statistical literature recognizes alternatives based on partially balanced incomplete block designs and optimal designs. Bose and Nair (1939) established the theoretical foundations for partially balanced structures, while John and Mitchell (1977) investigated the construction of incomplete block designs using statistical optimality criteria. Furthermore, Patterson and Williams (1976) developed alpha designs, demonstrating the usefulness of efficient incomplete block constructions for different experimental configurations.

On this basis, an extension of the design.bib() function is proposed, incorporating two design generation modes:

1. Exact balance: Construction of BIBDs that satisfy the corresponding combinatorial conditions.

2. Approximate balance: Construction of preferably equireplicated incomplete block designs, with pairwise treatment concurrence frequencies as uniform as possible, subject to constraints on the maximum number of replications per treatment.

The introduction of the r.max argument allows restrictions to be imposed on experimental size, while the approximate balance mode expands the range of feasible designs when exact balance requires an excessive number of blocks.

The evaluation of the resulting designs should consider connectivity, the distribution of pairwise concurrences, and statistical efficiency criteria, with the aim of identifying solutions that provide an appropriate balance between experimental precision and resource availability.

References

1. Bailey, R. A. (2008). Design of comparative experiments. Cambridge University Press. Chapter 11: Incomplete-block designs, pp. 219-240. doi.org/10.1017/CBO9780511611483.012

2. John, J. A., and Mitchell, T. J. (1977). Optimal incomplete block designs. Journal of the Royal Statistical Society: Series B (Methodological), 39(1), 39-43. doi.org/10.1111/j.2517-6161.1977.tb01602.x

3. Patterson, H. D., and Williams, E. R. (1976). A new class of resolvable incomplete block designs. Biometrika, 63(1), 83-92. doi.org/10.1093/biomet/63.1.83

4. John, J. A., and Williams, E. R. (1995). Cyclic and computer generated designs (2nd ed.). Chapman and Hall.

5. Bose, R. C., and Nair, K. R. (1939). Partially balanced incomplete block designs. Sankhya: The Indian Journal of Statistics (1933-1960), 4(3), 337-372.

See Also

design.ab, design.alpha,design.split, design.crd , design.cyclic , design.dau , design.graeco, design.lattice, design.lsd, design.rcbd, design.strip

Examples

Run this code
library(agricolae)
# out1<-design.bib(1:4, 3, balance = "exact", seed=101)
# out2<-design.bib(1:8, 5, balance = "exact", r.max=35)
# out3<-design.bib(1:8, 5, balance = "approximate", r.max = 5)
# 4 treatments and k=3 size block
trt<-c("A","B","C","D")
k<-3
outdesign<-design.bib(trt,k,serie=2,seed =41,kinds ="Super-Duper") # seed = 41
print(outdesign$parameters)
book<-outdesign$book
plots <-as.numeric(book[,1])
matrix(plots,byrow=TRUE,ncol=k)
print(outdesign$sketch)
# write in hard disk
# write.csv(book,"book.csv", row.names=FALSE)
# file.show("book.csv")

Run the code above in your browser using DataLab