Terminology

• Response Variable (Dependent Variable):

  Outcome of an experiment; a measurement

  Examples:

  • Response time for workstation

  • Time to complete a task for user interface

• Factors (Independent Variables):

  Variables whose values are chosen by the experimenter and that affect response variable

  Examples:

  • CPU type

  • menu style

• Levels:

  Values that a factor can assume

  Example: CPU type has three levels: Pentium, 80486, or 80386

• Primary Factors:

  Factors whose effects need to be quantified

  Example:

  • Workstation: CPU type, memory size, number of disks

  • User Interface: Organization, Menu Structure, Experience

• Secondary Factors:

  Factors that impact response variable but whose impact we are not interested in quantifying

  Example: Factors not listed under "Primary Factors"

• Replication:

  Repetition of some or all experiments; if all experiments are repeated three times, the study is said to have three replications.

• Design:

  Specification of number of experiments, factor level combination for each experiment, number of replications.

  Example: Trying all combinations of factors in workstation study requires 3x3x4x3x3 = 324 experiments.

• Experimental Unit:

  Any entity that is used for the experiment -- a human subject

• Interaction:

  Two factors A and B are said to interact if the effect of one depends on the level of the other

  Throughput in MIPS (millions of instructions per second)

  	|	      Factor A
  Factor B|	4M		16M
  --------+--------------------------------
  80386	|	 5		 7
  80486	|	14		16
  

  In table above, no interaction: MIPS grows by 2 regardless of processor, and by 9 regardless of memory size.

• Interaction graphs:

  Visualize factor interactions with an interaction graphs. Construct two graphs to represent one table:

  y-axis:

  response variable

  x-axis:

  one factor

  Each curve:

  the other factor

Example of no Interaction

	|	      Factor A
Factor B|	4M		16M
--------+--------------------------------
80386	|	 5		 7
80486	|	14		16

Lines are parallel in both graphs.

Example of Interaction

	|	      Factor A
Factor B|	4M		16M
--------+--------------------------------
80386	|	 5		12
80486	|	14		16

Lines are not parallel in either graph. The 80386 performance improvement is more dramatic than that of the 80486 when memory is increased from 4M to 16M. Thus processor type and memory speed interact!