Blocking and covariants

Question 1

Why do we randomise?

Answer 1

to compare how groups react to a certain intervention

Question 2

What is non-random allocation bias?

Answer 2

may be a difference between treatment groups that isn’t due to chance alone
eg
maybe differences in people who want to try new/old drug

Question 3

How can bias be avoided when allocating treatment to experimental units?

Answer 3

using random allocation

eg balancing disease severity across groups as this will effect results

Question 4

what is simple randomisation?

Answer 4

each participant has an equal chance of being allocated to each treatment

how due to random chance may not produce balanced groups

Question 5

When would block design need to be used?

Answer 5

trail with 2 interventions (drug, placebo), disease severity can be predictive of outcome so must be evenly distributed across groups,

Question 6

what is block design (randomisation)?

Answer 6

allocating in blocks (multiple of treatment arm number) to evenly distribute the groups

permutations: AABB, ABAB ect ect each block is selected randomly and allocated

Question 7

When is blocking an appropriate method of randomisation?

Answer 7

if there are a small amount of characteristics & small number of strata

gets too complicated with 3 participant characteristics (27 combinations)

Question 8

what does having more interventions than block size cause

Answer 8

incomplete block design and statistical problems

Question 9

What are block usually defined by

Answer 9

age, sex or disease severity

spatial area

Question 10

Why use blocking?

Answer 10

to promote balance of characteristics across experimental units

useful when characteristics are a predictor of outcome

or when splitting an area into small subsections