Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

Chapter 1

Fundamentals of Experimental Design

When we begin to study experimental design, we have to look at two of the most basic types
of plans: observational and experimental.

1.1 Observational Plans


In an observational plan, we simply observe the objects or subjects. We generally want to
measure an effect and with this plan, we ‘”take what we get.” As a result of this, we often
do not get a random sample (r.s.). These plans are often very cheap and easy to conduct;
they are generally used when it is impractical/unethical to impose a treatment.

Example 1.1.1. Suppose we want to conduct a study on smokers. We cannot ask random
people to smoke and so we must use an observational plan.

Example 1.1.2. Recovery time for knee surgery —we cannot ask random people to get
injured

1.2 Experimental Plans


In an experimental plan, treatments are imposed deliberately on objects/subjects where we,
as the researchers can change the level of treatment. We are often looking to measure an
effect. These types of plans are more structure and thus are more costly.

Example 1.2.1. Oral Insulin: Suppose we want to compare the effects of taking oral
insulin daily vs no insulin at all. We can do this through an experimental plan.

1.3 Factors & Responses


Definition 1.3.1. Factor: An explanatory variable (e.g. an input to a model) that will
be changed or set on each unit in a study. The units are referred to as subjects when the
unit is a person; more generally, we consider units as objects

Definition 1.3.2. Factor Levels: Set of values applied to a factor. In our example 1.2.1
(insulin example), we could have considered the amount of insulin given as a factor level

1
2 CHAPTER 1. FUNDAMENTALS OF EXPERIMENTAL DESIGN

Definition 1.3.3. Quantitative Factors: Factors to which we can assign a meaningful


value (amount of insulin)

Definition 1.3.4. Qualitative Factors: Factors where we cannot attach a meaningful


value. These are generally descriptive in nature (gender, eye colour, hair colour, etc)

Types of Response Variates (Outputs)


1. Continuous: Real numbers; generally things which we measure (height, weight, age,
etc)
2. Ordinal: Some sort of order or rank (small, med, large, etc)
3. Count: Number of stanley cups the leafs have won in the last 48 years
4. Binary: Yes/No (did someone survive the past 5 years, win/loss, etc)

1.4 Fundamentals of Experimental Plans


Definition 1.4.1. Blocking: Refers to grouping objects with “like” characteristics to-
gether. We could block our objects based on age, gender, height, etc. Blocking helps
prevent confounding due to the variable that is being fixed. If the blocked variate has an
impact on the response, this can lead to more precise conclusions. Blocking may not be
possible due to cost; could also lead to carryover effects.

Definition 1.4.2. Replication: The idea of replication is to apply the treatment to as


many units as possible. This provides us with more confirmation and less sample (chance)
errors, however may not be possible due to cost.

Definition 1.4.3. Random Assignment: We can randomly assign treatments to unit


(think of names in a hand —reach in and randomly select names without replacement. We
select without replacement since we generally do not want to select the same unit twice
since selecting the same unit twice provides no additional information.

You might also like