Chemometrics - Avoiding pitfalls

Know your subject as well as you can.
- At very least, have very close access to someone who does.
- The decision processes in the inquiry phase require at least as much chemical/technical input as statistical knowledge.

Monitor characterized systems for changes.
- The results of your analysis assume that unmeasured variables are constant or unimportant. This can cease to be true at any time after your analysis. For instance, a change in blending techniques or proportions might invalidate future whiskey results.
Checks are needed to guarantee continued model validity.
This is especially true in industry.

Don't become too fond of any one technique.
- Each technique has strengths and weaknesses.
- The use of complementary techniques minimizes weaknesses.
- Overreliance on any technique makes the process more vulnerable to mistakes.