Methods of experimental control

1
Methods of experimental control

• Randomization and Matching
• Counterbalancing
• Controlling participant and experimenter error

2
Randomization

• Random selection a representative sample
• Basis of generalizability or external validity
• Exhaustive sampling is good, but difficult
• Random assignment equivalent groups
• Basis of controlling most evars
• Chance of nonequivalent groups covered in error
  term of statistical analysis
• Random sequencing order effects

3
Matching

• Since randomization assigns individual
  differences to the error term, it has one
  unfortunate effect reducing power.
• Matching techniques increase group equivalence,
  so it increases power
• Matching must be complemented with random
  assignment to produce randomized blocks designs
  or matched groups designs.

4
Matching tactics

• Hold extraneous variable constant
• Limits population size
• Limits generality of results
• Manipulate extraneous variable in design
• Beware whynot variables
• Use a yoked control design
• Equate participants Precision control
• Equate participants Frequency distribution

5
Counterbalancing

• Sequencing effects
• Order effect
• Carry-over effect
• Counterbalancing tactics
• Intrasubject for repeated measures ABBA
• Intrasubject/intragroup ABBA/BAAB
• Intragroup
• Incomplete Latin and Greco-Latin squares

6
Latin Squares

• In Latin Square counterbalancing, each different
  treatment is symbolized by a letter, different
  applications of treatments are in columns, and
  different occasions are in rows.
• The number of rows and columns equals the number
  of treatments.
• Each treatment letter appears once and only once
  in each row and in each column.

7
Latin squares

• A B C A B C D A B C D
• B C A B C D A B D A C
• C A B C D A B C A D B
• D A B C D C B A

8
Controlling experimenter and participant effects

• Double-blind placebo model
• Disguise techniques
• Retrospective report, probes, and the think-aloud
  technique Reactive effects caveat.
• Caution
• Blind techiques
• Automation