Slide 7

Question 1

What is the definition of Nuisance Variance?

Answer 1

Any source of variability which might affect the response, other than the conditions of interest

Question 2

What are the effects of Nuisance Variance?

Answer 2

Might Bias Results

Can reduce precision of comparison of interest

Question 3

What are two ways to handle nuisance variance?

Answer 3

1. Randomize over nuisance variable, turning it into chance error
2. Use blocking to turn nuisance variance into factor of design

Question 4

What are the positive and negative effects of Randomizing over the nuisance variable, turning it into chance error?

Answer 4

• Positive
  
  • reduces potential for bias
• Negative
  
  • Makes chance errors large
  • Makes difference, or “effect” harder ot detect

Question 5

What are the steps to using blocking to turn nuisance variance into a factor of the design?

Answer 5

• Break material into uniform batches, then run a bunch of little completely randomized experiments, combining results.
• Blocking controls for impact of nuisance variable so chance errors are small

Question 6

How is a CB[1] set up?

Answer 6

1. Sort units into blocks of similar units. The number of units in each block must equal the number of treatments (or treatment combinations)
2. Separately within each block, randomly assign treatments to units in a block - each unit gets one treatment.

Question 7

What are the three ways to get blocks?

Answer 7

1. Sort units into blocks
2. subdivide larger chunks of material into sets of smaller pieces
3. Reuse subjects or chunks of material (blocks) each of several time slots (which are the units).

Question 8

How is the latin squares method structured?

Answer 8

1. There are two nuisance factors and one (possibly compound) factor of interest.
2. The number of levels is the same for all three of these factors.
3. Every pair of factors is crossed: for each choice of two factors, every combination of levels of the two factors occurs exactly once.

Question 9

When is the best time to use the latin squares design?

Answer 9

1. Subjects exhibit a lot of variability
2. The order of the conditions has a systematic effect
3. The factor of interest is experimental

Question 10

How can we block the design if subjects exhibit a lot of variability?

Answer 10

Make subjects a nuisance factor in the design, if it is possible ot measure each subject under each of the conditions.

Question 11

How can we manage the experiment if the order of the conditions has a systematic effect?

Answer 11

Make time periods a nuisance factor in the design, and use a latin square to ensure balance

Question 12

How can we manage the experiment if the factor of interest is experimental?

Answer 12

You must be able to assign treatments to fit the pattern of the latin square.

Question 13

What is the structure of the SP/RM design?

Answer 13

1. Has experimental units of two different sizes. The larger units are blocks of smaller units.
2. Has two sets of treatments or treatment combinations. One set of treatments gets assigned to larger units as in a CR desing. The other set of treatments gets assigned to smaller units separately with each block, as in a CB design.