hassediagrams: generates the layout structure and restricted layout structure of experimental designs

Damianos Michaelides, Simon Bate, Marion Chatfield

Overview

The hassediagrams package provides tools to visualize the structure of experimental designs using Hasse diagrams. The package determines the structure of the design, summarised by the layout structure, and uses this structure to generate a Hasse diagram. This diagram describes the structure of the design and the relationships between the factors that define the design. By considering the randomisation performed, in conjunction with the layout structure, a list of randomisation objects can be identified, known as the restricted layout structure. The package can also be used to generate a Hasse Diagram of this restricted layout structure. Objects in the restricted layout structure can be used to identify the terms to include in the statistical model.

The package is an implementation of the methodology described in Bate and Chatfield (2016a and 2016b).

Bate S.T., Chatfield M.J. (2016a). “Identifying the structure of the experimental design.” Journal of Quality Technology, 48(4), 343–364.

Bate S.T., Chatfield M.J. (2016b). “Using the structure of the experimental design and the randomization to construct a mixed model.” Journal of Quality Technology, 48(4), 365–387.

Features

Functions

hasselayout()

Creates a Hasse diagram of the layout structure. To generate the diagram all this is needed is a dataset consisting of the experimental factors that define the experimental design.

itemlist()

A function needed to generate the objects required to generate the Hasse diagram of the restricted layout structure

hasserls()

Creates the Hasse diagram of the restricted layout structure. Inputs to the function are the rls object, generated using the itemlist() function, an object that defines the randomisation objects and an object that defines the randomisation arrows.

Installation

Install from GitHub (Development Version)

The development version of hassediagrams can be installed with:

# Install devtools (if not already installed)
install.packages("devtools")

# Install hassediagrams from GitHub
devtools::install_github("GSK-Biostatistics/hassediagrams")

Install from CRAN (when made available)

install.packages("hassediagrams")

Usage

Load the package:

library(hassediagrams)

Example: Generate Hasse Diagrams of the Layout Structure and Restricted Layout Structure

A fractional factorial design for investigating asphalt concrete production

Generate the Hasse diagram of the layout structure

hasselayout(datadesign=concrete, larger.fontlabelmultiplier=1.6, 
            smaller.fontlabelmultiplier=1.3, table.out="Y")

Generate objects required to create the Hasse diagram of the restricted layout structure

concrete_objects <- itemlist(datadesign=concrete) 

Define the names of the randomisation objects corresponding to the structural objects

concrete_rls <- concrete_objects$TransferObject 
concrete_rls[,2] <- concrete_rls[,1]
concrete_rls[27,2] <- c("AC^AG^CC^CoT^CuT \u2192 Run")

Generate the Hasse diagram of the restricted layout structure

hasserls(object=concrete_objects, randomisation.objects=concrete_rls, 
         larger.fontlabelmultiplier=1.6, smaller.fontlabelmultiplier=1.3, 
         table.out="Y", arrow.pos=8)

Documentation

For an introduction to the methodology, check the package vignette:

vignette("Introduction_to_hassediagrams")

and for an introduction to the package implementation visit the documentation.

Maintenance

This package is in a stable state and will only be updated for bug fixes.

The person responsible for monitoring this package is Simon Bate simon.t.bate@gsk.com.