Stata is doing nothing unilaterally here. The author
of the program, Tony Brady, wrote this in his program,
which is pretty transparent.
if `AFhat' < 0 {
scalar `AFse' = .
scalar `cilo' = .
scalar `cihi' = .
}
In short, the programmer, evidently, decided that negative fractions
make no sense, so that saying more about them is futile.
If you have an alternative point of view, clone the program
and change the code. That's the beauty of open source.
Nick
n.j.cox@durham.ac.uk
Marc Campo
> I have calculated several attributable fractions.
> When the AF comes up as negative however, Stata does
> not display a standard error and confidence interval.
> Why?
>
> . logistic incident_lowback age gender hours_week
> second_job strain xfer posi manual
>
>
> incident_l~k Odds Ratio Std. Err. z P>z
> [95% Conf.
> Interval]
>
> age .9849692 .0149313 -1.00 0.318
> .9561349 1.014673
> gender .9118571 .3196976 -0.26 0.792
> .4586644 1.812836
> hours_week 1.000273 .0169735 0.02 0.987 .9675525
> 1.034099
> second_job 1.0135 .3644665 0.04 0.970 .5008663
> 2.050812
> strain 2.311378 .7130899 2.72 0.007
> 1.262595 4.231337
> xfer 1.658223 .4354864 1.93 0.054
> .9910564 2.774517
> posi .7943614 .2137511 -0.86 0.392
> .4687834 1.346059
> manual 1.025606 .0583063 0.44 0.657
> .9174643 1.146494
>
>
> . aflogit posi, ref(posi=1)
>
> Term Ref. A.F. s.e. [95% Conf. Int.]*
>
> posi 1 -0.2742 . . .
>
> TOTAL -0.2742 . . .
>
>
> Any thoughts? Why can't you have a negative AF?
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/