Guesswork follows.
Nearby numbers could easily round to
different multiples of 0.000001.
Why not go with
local equal = (r(P_p1) - r(P_n0)) < 1e-6
Also, watch out for binary-decimal precision
issues. A power of 2 might be practical.
Nick
[email protected]
Hans J. Baumgartner
>
> I would like the prediction after a probit to be equal for 1 and 0
> classified.
> Thus, I have to adjust the cutoff(x). But which value should x have?
> The following loop should do the job, but somehow I have a
> problem with
> the rounding. I.e. the thing flips between the first two sub-loops.
> Can somebody help me on this?
>
> local cutoff = 0.33 /* starting value */
> forvalue x = 1/100000 {
> display `x'
> qui lstat, cutoff(`cutoff')
> local equal = round(r(P_p1),0.000001) -
> round(r(P_n0),000001)
> /* Sensitivity - Specificity */
> display `equal'
> if `equal' > 0 {
> local cutoff = `cutoff' + 0.000001
> display `cutoff'
> }
> else if `equal' < 0 {
> local cutoff = `cutoff' - 0.000001
> display `cutoff'
> }
>
> else if `equal' == round(0,0.000001) {
> lstat, cutoff(`cuttoff')
> }
> }
>
*
* 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/
