Computing Effects...

Recall the model for a 2**2 design:
     y = q_0 + q_a * x_a + q_b * x_b + q_ab * x_a * x_b

Recall that the q's are called effects.

Substituting the four observations of y, denoted y_1, ...,y_4, in a 2**2 experiment yields:

     
     y_1 = q_0 - q_a - q_b + q_ab = 15
     y_2 = q_0 + q_a - q_b - q_ab = 45
     y_3 = q_0 - q_a + q_b - q_ab = 25
     y_4 = q_0 + q_a + q_b + q_ab = 75

Solving for q's yields:

     q_0  = 1/4 ( y_1 + y_2 + y_3 + y_4 )
     q_A  = 1/4 (-y_1 + y_2 - y_3 + y_4 )
     q_B  = 1/4 (-y_1 - y_2 + y_3 + y_4 )
     q_AB = 1/4 ( y_1 - y_2 - y_3 + y_4 )

An equivalent representation is a sign table:

Experiment |    I     A     B     AB     y 
-----------+-------------------------------
1          |    1    -1    -1      1     15
2          |    1     1    -1     -1     45
3          |    1    -1     1     -1     25
4          |    1     1     1      1     75
-----------+-------------------------------
Totals:    |  160    80    40     20
Totals/4:  |   40    20    10      5 

Note: The last line contains the same q's we obtained before:

    q_0  = 40
    q_a  = 20
    q_b  = 10
    q_ab = 5