Replication

1
Random Assignment Randomization?
10/16/08 This discussion and my chalkboard work
took an hour!
Yes
No
Yes
Designed Experiment
Observational Study
Replication?
Assessment Study
Impact Assessment or Case Study
No
(From Skalski and Robson 1995)
2
What does random mean?
random, randomly, randomize
Random impetuous and haphazard movement without
careful choice, aim, plan, etc. haphazardly
3
Randomization
What do we mean by randomization?

• Mix up the order
• Cant repeat what we did
• See no pattern
• Cant explain what we did

4
Randomization
What do we mean by randomization?

• Randomize to place in random order

5
Randomization
What do we mean by randomization?

• Random To exist with a frequency defined by a
  probability distribution

Stochastic pertaining to, or arising from
chance involving probability random
6
Randomization

• What we do
• Randomly select pieces to sample
• randomly select random sampling
• Randomly assign a piece of material to a
  protocol
• randomly assign experiment
• Order items or protocols randomly
• randomly order
• Physically place items randomly
• randomly placed

7
Types of Randomization

• Each replicate unit has a known chance of being
  assigned to a treatment.
• Or
• Each sample has a known chance of being sampled
• The process is definable and repeatable.

Randomization ensures that the effects we observe
are reasonably believed to be true for the whole
set were interested in, not just for the subset.
8
Random Assignment
Controlling the assignment of individuals to
groups can insure no systematic difference
between groups.
Fertilizer A
Fertilizer B
Fertilizer A
Fertilizer B
9
Random selection of stands to sample

• Randomly select 8 stands to sample, out of 20
  possible stands
• Number all the stands consecutively, 1 to 20
• Use a random number table or a good random number
  generator (not Excel) to generate a random
  ordering of the pot numbers.
• Take samples in the first 8 stands in the random
  ordering.

10
Observational Study
Random Assignment distinguishes Designed
Experiments from Observational Studies

• Measurements taken on units that already exist in
  groups or with specific characteristics.

Designed Experiment

• Measurements taken on units that are assigned to
  groups or characteristics by investigator.
• Assignment is done so that there is no
  systematic way that units are assigned to
  treatments

11

• The function of randomization is to ensure that
  we have a valid or unbiased estimate of
  experimental error and of treatment means and the
  differences among the means.
  
• Randomization ensures that a particular
  treatment will not be consistently favored or
  handicapped in successive replications by some
  extraneous sources of variation, known or
  unknown. Steele and Torrie 1997
• Randomization is somewhat analogous to
  insurance, in that it is a precaution against
  disturbances that may or may not occur, and that
  may or may not be serious if they do occur.
  Cochran and Cox 1957

12
Inferences

• Observational studies can only report
  associations between responses and groups.
• Because you dont know, and cant be sure, that
  something unknown is responsible for the
  difference you see between your groups.
• Well-Designed Experiments allow you to draw cause
  and effect conclusions.
• Because in theory, all other effects known to
  affect the response have been controlled for
  through the design and random assignment.

13
Inference

• Note Natural resource studies are commonly a
  mix of observational and experimental studies.
  It is not easy to have an natural resource study
  that can make cause and effect conclusions!

14
Replication
Before we accept the existence of an effect, the
effect must be observable in replicates that
represent the range of variation over which
inference is to be made.
-Hurlbert (1984)
15
Extending the results of your study
Based on the results of your study Only to the
strict scope of inference Based on your
scientific expertise Maybe farther - but not as
a result of the study