search

fsolt / dcpo_dem_mood

1 stars 0 forks source link

analysis: add models 2-4 #9

Closed fsolt closed 3 years ago

fsolt commented 4 years ago
sammo3182 commented 4 years ago

Getting computational errors when running the GMM by regime type 

Error in solve.default(crossprod(WX, t(crossprod(WX, A1)))) : 
  Lapack routine dgesv: system is exactly singular: U[3,3] = 0

The error rose at the step of gmm estimation: https://github.com/fsolt/dcpo_dem_mood/blob/af81b264f9c4d24462bbaab19d21f33a49debd5a/paper/dcpo_demsupport.Rmd#L341-L349

It looks like that adding the uncertainties brings both more accurate estimates and more difficulty of identification.

You can replicate the error by running https://github.com/fsolt/dcpo_dem_mood/blob/af81b264f9c4d24462bbaab19d21f33a49debd5a/paper/dcpo_demsupport.Rmd#L402

@fsolt

sammo3182 commented 4 years ago

I bet, this plot provides better evidence to show the reviewers that it's not the efficiency issue causing the null results but Claassen missed something important.

image

On the other hand, shame on me but I can't figure out how to draw the ideal version of it with facet. There's always a NA column drawn automatically and unremovably. (See my desperate commit e58eb58c58f9d7c570f4de9a286758c1b34493d6)