katrina-slm-impacts.R
Opened this issue · 7 comments
> tabdi$INLASDM<-(inla.zmarginal(trIrhoWbsdm, TRUE)$mean * sdmm1$summary.random$idx[idxbeta,"mean"]+inla.zmarginal(trIrhoWgsdm, TRUE)$mean * sdmm1$sum .... [TRUNCATED]
Error in splinefun(xnew, xx, method = "hyman") : zero non-NA points
In addition: Warning message:
In sqrt(dx) : NaNs produced
> traceback()
5: stop("zero non-NA points")
4: splinefun(xnew, xx, method = "hyman")
3: inla.smarginal(marginal)
2: inla.emarginal(function(xx) c(xx, xx^2), marginal)
1: inla.zmarginal(trIrhoWbsdm, TRUE)
Did something change in the internals?
> o <- inla.zmarginal(trIrhoWbsdm, TRUE)$mean
Error in splinefun(xnew, xx, method = "hyman") : zero non-NA points
In addition: Warning message:
In sqrt(dx) : NaNs produced
> o <- inla.zmarginal(trIrhoWgsdm, TRUE)$mean
Error in splinefun(xnew, xx, method = "hyman") : zero non-NA points
In addition: Warning message:
In sqrt(dx) : NaNs produced
katrina-slm.R and katrina-slm-neigh.R run OK, it's just katrina-slm-impacts.R that doesn't complete.
No, I usually ignore branches and never use them. I always assume that branches were just for local work, and are on the central repo for reference only. I've been working only in the main branch of my fork - how should I merge your update with my fork main, or update my fork of your update to my main and your update ????
I'm trying to merge manually locally. Do not use your update branch, I will make a PR against your main branch.
I'm some way through. For boston, the use of eigenvalues assumes that n is small enough, but for house, the new impacts approach is not feasible. I'll PR to main branch only when completed, forget any other branches. I have no sarp.RData, will try to reconstruct from the spatialprobit package.