Week 1 Flashcards
Defining difference between experiment and observation
An experiment is defined by an intervention in the system
Define treatments
The distinct states being tested
(Experimental) units
The entities to which treatments are applied
Observational units
Entities on which observations are carried out
Degrees of Freedom
Number of independent comparisons that can be made
Resource equation and constituents
T + B + E =n-1
t: treatment - number of treatments -1
b: blocking - number of blocks - 1
e: error - used to estimate σ^2
Then increase error by adding until reaching 10
Main treatment structures
Unstructured
Levels of Quantitative variable
Factorial
Types of factorial
Incomplete
Complete
One way ANOVA model
aka full treatment model
τ_r (i) is difference between expected response form administering treatment and overall expected response
Properties of error in ANOVA
Expectation = 0
Var = σ^2
Main effect of a factor
Blocking
A group of ‘similar’ units should include roughly the same proportion of units for each treatment
Randomised complete block design
Each block contains a single replicate of each treatment
How many DoF should you have for a good estimate of error
At least 10
Interaction between 2 factors
You do not have to create blocks based on
Just one factor
Eg:
Normal equations
Derived in finding LSE of linear model
When partial derivs of SSE wrt each param must equal 0
