matrix D in quadratic function is not positive definite

[kr3118]

Hi,

I have tried with 2 datasets with both the meow extension and with just MAMMaL, and both times I get the error Error in solve.QP(Dmat = Sigma * 2, dvec = rep(0, ntaxa), Amat = Amat, : matrix D in quadratic function is not positive definite! Calls: CalculateWeightsE -> OptimizeWeights -> solve.QP. The full command line + log is as below:

./meow -s data.phy -c 60 -p K
Running MEOW with the following parameters:
Sequence file: data.phy
Output file: esmodel.nex (Total number of classes: 120)
Partition mode: K_eff
High partition classes: 60 Low partition classes: 60
Quantile: 0.75 Cluster type: hclust Penalty: 1e-10

Number of sequences:1477; Number of sites:580
Processing sequence file...
Detected 108 invariant sites
Creating high and low partitions...
Calculating likelihood weights...
Error in solve.QP(Dmat = Sigma * 2, dvec = rep(0, ntaxa), Amat = Amat, :
matrix D in quadratic function is not positive definite!
Calls: CalculateWeightsE -> OptimizeWeights -> solve.QP
Execution halted

How do I get around this?

Best,
Kenta

[kr3118]

I managed to get it to work when I disable likelihood weighting... How much will this impact the resulting profiles?