[Previous Months][Date Index][Thread Index][Join - Register][Login]
  [Message Prev][Message Next][Thread Prev][Thread Next]

[IP] RE: Odds

Michael [mailto:email @ redacted
<mailto:email @ redacted> ] wrote:

> heh... heh... your employer (indirectly) See the FAQ page near the 
> bottom for the essence of the papers and links to the original and 
> NIDNK(sp?)

OK - those are risks, not odds.  Risk in this case is a technical
statistical term.  It is calculated as incident cases in one group divided
by its respective population divided by the incident cases in another group
divided by its population.  IOW, risk is meaningless unless there is a
comparison group, either implied or explicit.  There is a lot of room for
confusion because there are multiple scenarios on which the comparisons
could be based.

Not to slam my colleagues at NIH, the table is more confusing because it is
based on prevalence, not incidence, so the number are more likely odds
ratios, not risks as labeled.  

Finally, the disease process in type 1 is sufficiently complex that a simple
statement of odds ratios really doesn't tell the story here (nor does it in
most health studies).  Stated another way, the late John Tukey once said
that there is no good statistical model.  Some are useful.

The only reason that I'm going on about this is the distinct possibility
that a couple may decide on having a family or not relying too much on a
very simplistic number in a table that is probably taken somewhat out of

Finally, none of this points to type 1 being sex linked as was implied in
the post I responded to.

James Handsfield, PhD, MPH
Centers for Disease Control and Prevention
email @ redacted
for HELP or to subscribe/unsubscribe, contact: