Hi,

I am trying to devise a LRT (or perhaps use the AIC difference) where the null model is neutral evolution and the alternate model is selection.

For example Yang's M1a vs M2a does this (thus allows use of chisquare with df = 2)
M1a (neutral):
parameters: p0, w_0 < 1, w_1 = 1
M2a (selection):
parameters: p0, p1, w_0 < 1, w_1 = 1, w_2 > 1

I was thinking about setting the neutral model (using the general discrete bi variate distribution) to be:
beta_s <= alpha_s for all s
And alternate model is:
beta_s > alpha_s for at least one s

Can you constrain these variables in hyphy?

Hi there,

The types of constraints you need can be implemented through ratios, e.g.

[code]

global R_0;

beta_0 := alpha_0 * R_0;

R_0 :< 1;

...

[/code]

Alternatively, you can use the ratio parameterization directly (e.g. see Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login)

Sergei