Statistics for Practical People

Part II - How to Get Statistical Help
Published August 1988

Most Biometricians and Statisticians aren't complete jerks, although I grant you they have their share. Most of them really want to be helpful. If they aren't, it's because they are inexperienced in dealing with clients or don't have a proper respect for the field work. I once heard that it takes 200,000 casualties to properly train a Major General. I suppose most biometricians have to personally kill or cripple at least 15 projects before they learn their craft. Herein lies the first lesson.
  • Look for some grey hairs. Let somebody else train the whiz kids. Remember - you are looking for some good advice, not tricky math. You want to use techniques which have a track record of good results, so find somebody who has been around the block a few times.
  • Tell him the REAL Problem. Many people only seek help when they are already hung up, usually on some mathematical point, and often they are trying to do something foolish anyway. They come to talk about the glitch in their solution - not to talk about their problem. This is deadly, and a well-meaning but inexperienced consultant will solve the glitch without ever solving the problem.

For example: A cruiser came to a biometrician with pages of notes concerning sample sizes, statistical tests, tracing of wood flows etc. A 6-month project for sure. After the fifth time he was asked, "What's the real problem here?", he bursts out "Well I'll tell you - there's this wet-behind-the-ears young graduate from _______ who has been writing snotty letters to the manager, who is jumping on my boss about something that flatly can't be done anyway! "The cruiser, of course, was quite right. All it took was a phone call to make the snotty letters go away, and the project and its cost vanished.

There are lots of people who just love the sight of a "Gordion Knot", but the best consultants will find a way to avoid most glitches rather than undo them with your time, money, and sweat. A neat swindle that simply solves the real problem may not generate grant money for a university or create an unreadable publication in a pretentious journal, but it's worth its weight in gold. Which brings us to another point.Next Column

  • Teach each other your craft. He should give you a statistics lesson, so that you are comfortable with where your problem fits into the statistics field. He should teach you the statistical jargon and what it means. He should cover the statistical options, and why he recommends a particular solution from that pack. He should be able to send you home with something a human being can read. For your part tell him about the history of the project, how it fits into the political situation and any options you might see. If you want him to keep his mouth shut about some of that information, tell him so. If it's a political problem, tell him that, so he doesn't waste time on non-productive options.
  • Do the groundwork. When you have a solution, don't let it die because of politics, personalities or administrative problems. As a cruiser supervisor once said to me: "If you have a prophet of doom in a high place, and he predicts failure - count on it!" Make sure that the solution has a chance of success. No point in casting seed on infertile soil. People who do so aren't known as hard working optimists - they're known as dummies.
  • Reward the Virtuous; Train the Rest. Good statistical advisors, like good computer programmers, are both rare and valuable. If you appreciated their work say so. Think about how to reward him. Write a "thank you" letter, with copies to his boss, and maybe, his boss' boss. Tell him why his approach was valuable. Everybody needs feedback.

If you don't get good advice, try to train him. If he didn't appreciate the field situation, get him out there, in poor weather if possible, and let him handle every piece of equipment and do every task. The good ones will welcome the chance. Tell him why his solution failed, without putting it in writing or involving his boss -- this is training....right?

Finally, if you get a good working relationship going, pass the word to your friends. Life is hard enough without the problems caused by bad statistical advice. With a bit of thought, you can avoid being a victim, and with some luck, it will be well worth your time to get advice as early in your project as possible.


Return to Home
Back to Contents