LETTER: Ethical code for statisticians

From Dr. R. P. Saundby Sir: Regarding your article `Lies and damn lies' (19 December), from 1987 to 1988, I was director of medical personnel and training in the Ministry of Defence. A project at the time was to restructure the pay of armed services medical officers to a form more in parallel with that of a career in the NHS.

Having produced a paper for Treasury consideration containing statistical information that had been cleared by the MoD, I was shocked to hear a senior civil servant pass it to his statistics department with the instruction, "See if you can challenge these figures." My view was that although the merits of any proposed change were properly subject to debate, the facts on which the case was to be argued should be established beyond doubt as a preliminary to the main discussion.

I therefore arranged a meeting between the statisticians of the Treasury, the MoD and the British Medical Association at which I proposed that there was a general ethical duty of statisticians to establish agreed figures, and that they should not agree to behave as destructive members of a partisan team. Decisions were more likely to be sound if based on the best data available. The provision of "best data" was a professional responsibility of statisticians and it was unethical to allow figures to be distorted by covert departmental aims.

This tactic was a great success. The statisticians were unanimous in their dislike of political pressure and their desire to treat statistical information in a neutral, dispassionate manner.

At the next Treasury meeting, the mandarin concerned repeated his tactic, only to be told by his own statistician: "These figures are accurate, and have been agreed by the statisticians of the parties concerned prior to this meeting."

I would suggest that you emphasise the ethical duty of professional statisticians to present data which is non-partisan. By this tactic you will be seen to be supporting a profession, rather than launching an attack on individuals who have usually been subject to unfair pressure.

Statisticians should be encouraged to develop an ethical responsibility so that none will support data that has been distorted. Failure to meet these standards should put professional careers at risk. I think that this would be welcomed by most responsible statisticians.

Yours faithfully, R. P. SAUNDBY Llangynidr, Crickhowell