Confoundedparameter problems can usually be avoided by thinking about the meaning of each parameter in the context of the data. If the information it conveys is already being supplied by an earlier parameter, leave it out. In the example below, a threestage process (with two transitions) can be modeled by adding two logistic models. Since a logistic model has four parameters (rate, center, height, and floor), one might expect eight parameters in the compound model. But are any of these parameters redundant?
Example 4: The data to the right record the depth of the cut of a milling machine during a portion of a production run. Fit an appropriate model to the data to make estimates of the time, to the nearest millisecond, of the midpoint of each transition.
Solution:
Step transitions can be modeled by logistic functions; for this data, the sum of two logistic functions would be suitable.
The horizontal asymptotes in a logistic graph are controlled by the floor and height parameters. But in this sumoflogistics model, floor2 (the floor of the second logistic) will equal floor1+height1, and does not need a separate parameter in the model.
In this case, it appears that the model can be further simplified by using the same height and rate parameters for both transitions, leaving the model with five parameters: floor, height, rate, center1 ,and center2, (use cells G3–G7 for these). 
Secs 
Depth 
20.0 
28.46 
20.1 
28.47 
20.2 
28.50 
20.3 
28.58 
20.4 
28.61 
20.5 
28.69 
20.6 
28.63 
20.7 
29.89 
20.8 
38.27 
20.9 
56.00 
21.0 
61.44 
21.1 
62.05 
21.2 
62.34 
21.3 
62.35 
21.4 
62.37 
21.5 
62.22 
21.6 
62.19 
21.7 
62.37 
21.8 
62.43 
21.9 
62.49 
22.0 
64.81 
22.1 
78.21 
22.2 
92.96 
22.3 
95.77 
22.4 
95.95 
22.5 
95.93 
22.6 
95.99 
22.7 
96.08 
22.8 
96.06 
22.9 
96.20 
23.0 
96.25 
