You are here

ACE model error: In model 'ACE' Optimizer returned a non-zero status code 6.

5 posts / 0 new
Last post
MAchterberg's picture
Offline
Joined: 02/20/2017 - 08:51
ACE model error: In model 'ACE' Optimizer returned a non-zero status code 6.
AttachmentSize
File Dataset17.71 KB
Microsoft Office document icon Script29.5 KB

Hello,

I am quite new to behavioral genetic modelling and I have a question about an error that I am receiving since the OpenMX update.
I have used the same scripts as before the update, but my results are very different.
in addition, I now get the error:
In model 'ACE' Optimizer returned a non-zero status code 6. The model does not satisfy the first-order optimality conditions to the required accuracy, and no improved point for the merit function could be found during the final linesearch (Mx status RED)

Does anyone else received this message since the update?
Is there a way to solve it?

My data and scripts are attached.
The "old" results were: A=0,55 C=0 E=0,45, AIC= 252,12

Many thanks in advance!
Best, Michelle

AdminRobK's picture
Offline
Joined: 01/24/2014 - 12:15
switch optimizers?

FTR, the message you're reporting is a warning, not an error. Anyhow, you might be able to avoid it if you switch optimizers. Put

mxOption(NULL,"Default optimizer","SLSQP")

in your script, after you load OpenMx.

AdminRobK's picture
Offline
Joined: 01/24/2014 - 12:15
SLSQP works for me

I tried running your script with SLSQP instead of CSOLNP (which is now the on-load default). I recovered the standardized point estimates and AIC you report in your post.

MAchterberg's picture
Offline
Joined: 02/20/2017 - 08:51
Thank you!

I added the mxOption to the script, and now it works fine!
Thank you so much :)
Best, Michelle

tbates's picture
Offline
Joined: 07/31/2009 - 14:25
starting values

Mike Neale pointed out off line that this is a starting value problem, where some of our optimisers can cope beginning a very long distance from the solution

Starting a, c, and e at 1/3 * sd is best. In multivariate space you'd start with only the diagonals of Cholesky matrices initialized at these values (and zero elsewhere).

#Figure out reasonable starting values
svMean <- colMeans(AmygdalaHPC[AmygdalaHPC$zygosity==1, 2:3])[1]
svPaths <- sqrt(var(AmygdalaHPC[AmygdalaHPC$zygosity==1, 2:3], na.rm=T)[1,1])/3