search

gaynorr / AlphaSimR

R package for breeding program simulations
https://gaynorr.github.io/AlphaSimR/
Other
43 stars 18 forks source link

Report if user-provided covariance is not positive definitely #98

Closed gregorgorjanc closed 1 year ago

gregorgorjanc commented 1 year ago

Is your feature request related to a problem? Please describe. Based on discussion here https://github.com/gaynorr/AlphaSimR/discussions/97#discussioncomment-4159623 it would be best to warn users when their covariance matrix is not positive definite, so they are warned that simulation outcome can be different to what they expect

gaynorr commented 1 year ago

This should already be the case. The following code should produce the warning below.

library(AlphaSimR)

founderPop = quickHaplo(100, 10, 100)

R = diag(3)
R[1,2] = R[2,1] = 0.9
R[1,3] = R[3,1] = 0.9

SP = SimParam$
  new(founderPop)$
  addTraitA(100, mean=rep(0,3), 
            var=rep(1,3), corA=R)$
  setVarE(h2=rep(0.3,3))

Warning message: In transMat(corr) : Matrix is not positive semi-definite, see ?transMat for details