|
In Reply to: Re: There is a slightly problematic problem posted by Passing By on May 12, 2005 at 22:33:21:
I'm willing to work with a different set of numbers. The kind of stats we're doing here is as much an art as a science. The main thing is to be very clear about the rationale for selecting one set of figures over another. If you were in a position to know the actual numbers of live-in, full-time membership in 1973 or 1969 or some point in the early history of the group, please share the information and I'll do the math.
I've been thinking about this problem some more, btw, and I think some confusion about calculating a suicide rate on 11/32,000 exmembers and arriving at ratios per 100,000 ppl is related to understanding the difference between incidence and prevalence.
Also, there is a secondary question being raised here, which is a desire to estimate fluxuations in TFI membership over time based on birth rates, new adult membership enrollment, and departures from the group.
It's one thing to estimate the membership population over time. We might want to do that, but it isn't necessary to do that to arrive at a suicide rate for the former membership at a particular point in time.
Prevalence represents the proportion of people alive on a certain day who previously had a diagnosis of a condition, regardless of how long ago the event occurred, or if the patient is still under treatment or is "cured".
Prevalence estimates require additional data to adjust for birth and death rates. But as you can see, suicide is not a condition that the person lives with or recovers from. Suicide is a subset of the death rate.
More on prevalence can be found at: http://srab.cancer.gov/prevalence/definitions.html
What we're trying to arrive at with a known suicide number (11 or 20) is incidence of an event in a known population at a given point in time. Incidence is the occurence of an event relative to something, such as a known population. Incidence is calculated as a rate of occurrence in a known population at a particular time of measurement, e.g., 11/32,000=.000344. The incidence rate is then applied to 100,000 ppl so that it can be compared to other known populations at a particular point in time.
Rates in the general population are very small and do not change appreciably over time. It takes a good 10 years to determine whether suicide rates are going up or down in a meaningful way.
I'm going to keep checking my methodology, because this stuff does get tricky, and I could be wrong about the way I'm thinking about the problem. I'm always open to constructive criticism and challenges.
