dr fisher’s casebook
TRANSCRIPT
feature
26 march2004
Consultant medical statisticians, unlike general practitio-
ners, get no real “on-the-job” training—from day one a cal-
low youth has to face grizzled medical practitioners with
years of experience of extracting P-values from statisticians.
Doctors, from housemen upwards, learn in their fi ve years
at medical school the subtle art of persuading other people
to do what they want, starting with nurses when they fi rst
arrive on the wards, and moving on to statisticians when
they embark on that dangerous career called research. In
statistics it is absolutely clear what they desire—P-values,
many of them, and invariably less than 0.05.
' at is about the limit of their knowledge of
statistics—to fi nd “signifi cance”. I have even had ur-
gent faxes from international conferences: “Please fax
P-values asap, as I need them for presentation tomor-
row”. I suppose it is our fault really; we should not have
appropriated that word “signifi cance”. It sounds, well,
important and so much to be desired by the medical fra-
ternity. If the founding fathers had called it “unlikeliness”
I doubt we would have the problem we have today.
Being a consultant statistician is a steep learn-
ing curve, as they say nowadays, and many an enthu-
siastic junior statistician has sunk under the weight of
responsibility of being the only person in a medical de-
partment who has a passing acquaintance with num-
bers. It is a daunting responsibility to be handed sheets of
scribbled numbers (or these days a disk containing an Ex-
cel spreadsheet) and to know that publication and the fame
and prosperity of the investigator depend on your being able
to wring some signifi cance (in the non-statistical meaning
of the word) out of the data, and that many thousands of
pounds have been spent in generating these numbers.
Much of one’s training will have been devoted to
testing a single hypothesis—how do you now cope
with potentially a thousand hypotheses generated by
an extensive and ill-thought-out questionnaire? Old
hands know that the answer is simple—much data
collected in the name of medical research can safely
be ignored—the years of experience, the scars from
abrasive personalities, all these contribute to one being
confi dent about which parts of the data can be ignored.
Pharmaceutical statisticians, of course, don’t have that
luxury—unless they produce more tables than there
are items of data they haven’t earned their (substan-
tial) salaries. ' e one clear advantage of the program
SAS is that it can produce reams of carefully format-
ted tables with little eff ort.
Generally, statisticians do consulting because they
genuinely want to help doctors—and there are few more
eager to help than the newly qualifi ed statistician. How-
ever, I think it highly undesirable to send junior statis-
ticians alone into a shark-infested medical department.
' ey need fi rst to learn to swim in a shoal with other
statisticians, where they can bounce ideas off seniors.
Only then will they learn what help they can most use-
fully give.
It is important to be recognised as one of the team
of scientists, not a back-room boy who does the num-
bers. To do this properly requires a basic understand-
ing of the culture of medical research, what is important
and what is less so, and this takes time.
It also helps to have a sense of humour. I was once
told by a colleague that he had been approached on a
statistical problem by a colleague. “Why don’t you go
and see Dr Fisher?” he asked. “Well,” came the reply,
“last time I told him my problem he burst out laugh-
ing!” I hadn’t thought I was being unkind. I believe all
I said was: “You want to prove what? With that size
sample?”
Dr Fisher has worked as a medical statistician for more years than are good for him. He has tried to teach statistics to a wide variety of people, including students in professions allied to medi-cine, medical students, registrars and post-graduate statisticians, in different countries and with varying degrees of success. He has also worked with people from a wide variety of medical speciali-ties, and feels he is the typical “jack-of-all-trades”. However, he doubts that any other occupation could give as much varied expe-rience and insight into medical practice.
Life as a jobbing statistician in a medical school bears a strong resemblance to life as a general practitioner: you have to be a jack-of-all-trades, you hold clinics where all the great unwashed can come, unfiltered and uneducated, and you get patronised by the consultants.
Dr Fisher’s casebook
"Get the stato to run the data
through the computer"