Example of Two Variables---III

• Unfortunately, there is no reason to believe that this large shift is well estimated as assumptions that lead to expansion break down.

  Terms like are quite likely to be big.

• We see that estimate in Equation (30) is false and one will typically jump out of valley and find

  

• The method will diverge!

• In the above example, the estimate for was valid and not spoiled by estimate.

• In a real example, these will be mixing between and due to nonzero .

• Both shifts , will be large and specious.