Re: What to do with "log of a zero rate"?

On Jan 24, 5:13 pm, z <gzuck...@xxxxxxxxxxxxxx> wrote:
On Jan 24, 4:10 pm, Bruce Weaver <bwea...@xxxxxxxxxxxx> wrote:



On Jan 22, 11:36 am, Ray Koopman <koop...@xxxxxx> wrote:

On Jan 22, 7:17 am, JeanCK <svel...@xxxxxxxxx> wrote:

--- snip ---

Log[rate/(1-rate)] is for binomial-type rates in which the numerator can not exceed the denominator. But an employee can be injured more than once, so the rate can exceed 1.

Ray's comment got me thinking. In cases like this, I think
occupational health researchers may compute an accident density rather
than a rate. For accident density, the denominator would be some kind
of person-time variable (e.g., person-hours).

--
Bruce Weaver
bwea...@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx/wv/bwhomedir
"When all else fails, RTFM."

depends what you're trying to do, of course. maybe the question is
just, the percentage of workers who get injured, at all.


A former colleague who does occupational health research suggested
looking at Appendix B of the document available here:

http://www.iwh.on.ca/products/eval.php

It appears that the total number of hours worked is often used as the
denominator.

--
Bruce Weaver
bweaver@xxxxxxxxxxxx
www.angelfire.com/wv/bwhomedir
"When all else fails, RTFM."
.

Relevant Pages

  • Re: What to do with "log of a zero rate"?
    ... occupational health researchers may compute an accident density rather ... For accident density, the denominator would be some kind ... Bruce Weaver ... "When all else fails, RTFM." ...
    (sci.stat.consult)
  • Re: What to do with "log of a zero rate"?
    ... But an employee can be injured more than once, so the rate can exceed 1. ... For accident density, the denominator would be some kind ... For the second model there will still be the "problem" of the log, I'll try to use the hints already given. ...
    (sci.stat.consult)
  • Re: What to do with "log of a zero rate"?
    ... occupational health researchers may compute an accident density rather ... "When all else fails, RTFM." ... just, the percentage of workers who get injured, at all. ...
    (sci.stat.consult)
  • Re: Calculate Standard Deviation with R
    ... The SD and variance are computed with n-1 in the denominator. ... Bruce Weaver ...
    (sci.stat.math)