Abstract:
Conclusion: The first part of this note deals with problems of propagation of error and the latter part with questions commonly identified as "model validation" questions. It is concluded that all validation questions
can be resolved into two parts: (1) strategic aspects, involving logical and inferential structures and (2) tactical aspects, involving some "propagation of error" technique. It is also concluded that the questions commonly lumped under
"sensitivity analysis" can be divided into two distinct categories. One group involves propagation of error, and the other a description of model behavior. Although these are not unrelated, they are sufficiently
different operationally that the distinction seems desirable. The problem of too many variables is discussed and it is recognized
that a strategy to resolve this problem must be worked out. It is anticipated that this problem will he even,more critical in the study of model behavior and that the note on that subject must deal with the problem.
