What do people think about using a double cumulative normal for an asymptotic selectivity. This would allow a bit more flexibility than using the logistic or a half normal using the double normal.
It could simply be added under the double normal definition when the plateau is -999, for example.
See equation 2.8 in Julio, J.M., 2007. The Fan Chart …. Borradores de Economia No. 468
Let me know if you think this is a good idea.
S(i) = C*sqrt(2pi)*sd1*cumN(i,mean,sd1) if i<=mean
S(i) = 1- C*sqrt(2pi)*sd2*(1-cumN(I,mean,sd2)) if i>mean
C=sqrt(2/pi)/(sd1+sd2)
ADMB has a cumulative normal distribution
Let me know what you think,
Mark
--
You received this message because you are subscribed to the Google Groups "SS3 - Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to
ss3-forum+...@googlegroups.com.
To view this discussion visit
https://groups.google.com/d/msgid/ss3-forum/70d49261-d395-4e41-aed1-cd1c3af2aaf9n%40googlegroups.com.