<html><body><div style="color:#000; background-color:#fff; font-family:arial, helvetica, sans-serif;font-size:12pt"><div><span>
<div class="p1">Hello ADMB community,</div>
<div class="p2"><br></div>
<div class="p1">I am relatively new to modelling, especially with ADMB and I am currently trying to run a series of glmm on data using the beta distribution. The response variable are genetic admixture in fish and is measured as a proportion, the fixed variables are continuous environmental parameters and the random variable is the lake (17 different lakes) in which sampled fish came from.</div>
<div class="p1">I created several models with a single fixed variable each using the following syntax:</div>
<div class="p2"><br></div>
<div class="p1">glmm15<-glmmadmb(formula = Admix ~ StMonth + (1|Lake), data = data, family = "beta")</div>
<div class="p2"><br></div>
<div class="p1"> and they seem to work fine, however several parameters result in the following error:</div>
<div class="p2"><br></div>
<div class="p3">Error in glmmadmb(formula = Admix ~ StMonth + (1 | Lake), data = data, : </div>
<div class="p3"> The function maximizer failed</div>
<div class="p3">In addition: Warning message:</div>
<div class="p3">running command './glmmadmb -maxfn 500 -noinit -shess' had status 1</div>
<div class="p2"><br></div>
<div class="p1">I've tried running this with verbose=T and the message "matrix not pos definite in sparse choleski" appears in red within the analysis and just before the model finishes running there is the message "Hessian does not appear to be positive definite"</div>
<div class="p2"><br></div>
<div class="p1">I am currently using a mac running os X 10.6.8 with the most recent version of R (2.14.0) and the glmmADMB package which was available on December 9th 2011. </div>
<div class="p2"><br></div>
<div class="p1">I would greatly appreciate any advice or suggestions people may be able to provide regarding this problem.</div><div class="p1">I've attached a smaller version of my database with the random variable, response variable and three fixed parameters, one which seems to work (pH) and two which return error messages or warnings (SDI and StMonth)</div><div class="p1">I've also attached a test file with sample r codes.</div><div class="p1"><br></div><div class="p1">Sincerely,</div>
<div class="p2"><br></div>
<div class="p1">Andrew Harbicht</div></span></div></div></body></html>