Matrix concepts: Not positive definite, Hessian, Jacobian
This is a stub article: email maintainer("umx") if you’d like it finished.
What is the Hessian?
- The Hessian is the matrix of second partial derivatives of the fit function with respect to the parameters, and is the inverse of the covariance matrix of the parameters.
What is the Jacobian?
# list up manifests and latents
manifests = nfc_items
m2 <- umxRAM("NFCIQ", data = mxData(cov(nfc[,manifests], use = "pair"), type = "cov", numObs = nrow(nfc),
umxPath(var = manifests),
umxPath(v1m0 = "NFC"),
# Create NFC as a latent, loading on each NfC item
umxPath("NFC", to = manifests),
umxPath(facets, to = "NFC"),
# Let all the facets intercorrelate
umxPath(unique.bivariate = letters[1:3])
)
umxSummary(m2, show="std")
m2 = umxRAM("formative", data = mxData(theData, type = "cov", numObs = nrow(demoOneFactor),
umxLatent("G", formedBy = manifests, data = theData)
)
umxSummary(m2, show="std")
plot(m2)