Experimental Design All experiments consist of two basic
Experimental Design All experiments consist of two basic structures: Design structure Method of grouping EU’s into homogenous groups or blocks Treatment structure Consists of the set of treatments
Experimental Design Choosing the design and treatment structure Design structure - chosen using all available knowledge of the experimental material Set of treatments - determined by the objectives of the experiment.
Experimental Design Choice of experimental design is very important!
Experimental Design Primary objective when choosing a design structure: Reduction of experimental error Do this by grouping EU’s so that conditions as uniform as possible. Grouping must occur BEFORE treatments are assigned!
Experimental Design Types of Design Structures Completely Randomized (CR) Randomized Complete Block (RCB) Latin Square (LS)
Experimental Design Completely Randomised • • • Simplest and most powerful Treatments assigned randomly Randomisation not restricted in any way
Experimental Design EXAMPLE • • • 20 cows – same age, weight, breed Objectives of study: effects of vitamin supply Treatments: 4 different vitamins (A, B, C, D) D D A A C C C B B A A B B D D C A B C D
Experimental Design Randomised Complete Block • • • EU’s arranged in homogenous groups (= blocks) • Each block contains each treatment Grouping based on a single outside source of variability One restriction on randomisation Treatments assigned randomly to EU’s within a block Every treatment occurs same number of times within each block
Experimental Design EXAMPLE • • • 5 different soya bean cultivars (A – E) 3 different planting locations Source of variation = location = block Block 1 Block 3 A B C D E Block 2 A B C D E
Experimental Design Latin Square • Blocking occurs in two different ways, so two restrictions on randomisation. • • Each row and each column is a complete block Treatments assigned so each treatment once in each column, and once in each row
Experimental Design EXAMPLE Objective: To determine the effects of 4 different diets on liver cholesterol in sheep (A – D) Sources of variability: • • body weight 1 (groups 1 - 4) 2 Age 3 (groups 1 – 4) 4 1 2 3 4 A B C D B A D C A B
Experimental Design Treatment Structures Determined by objectives • Imposed (controllable) eg type of protein in diet • Uncontrollable eg gender of animal
Experimental Design Types of Treatments Define either: Unstructured populations - compare sets of unrelated, qualitative treatments eg yield of maize hybrids
Experimental Design Types of Treatments or: structured populations – compare related populations May have a group or gradient structure eg Diet contents or levels of exercise
Experimental Design Types of Treatment Structures Two most common: One-way treatment classification Factorial treatment arrangement
Experimental Design Types of Treatment Structures One-way treatment classification • Simplest type • Unstructured or structured • Observations – based on one set
Experimental Design Types of Treatment Structures Factorial treatment arrangement • Set of treatment combinations • Qualitative or quantitative • Each factor occurs in combination with every other factor • Provide information about interactions among factors
Experimental Design Planning and pre-experiment protocol • Obtain unbiased estimates of treatment effects • Estimated with adequate precision to detect differences • • Max information from resources Protect against erroneous conclusions
Experimental Design Planning and pre-experiment protocol Preliminary questions: • • • What is to be accomplished? What variables will be measured? What is the population involved? How many treatments and how arranged? Controls to be included?
Experimental Design Planning and pre-experiment protocol Preliminary questions cont: • • • What are the EU’s? What is the design structure? Replication? Variability among EU’s? Type of data? Can the desired comparisons be made?
- Slides: 20
