haziqj / lavaan.bingof Goto Github PK

The current bottleneck in the test statistics calculation is the call to lav_model_vcov from the {lavaan} package. It is more apparent when $p$ is large (e.g. $p=15$ models).

profvis results

• print() can display the simulation settings.

• summary() would do something like the mean rejection rate. Needs input sig_level = 0.05.

Need to also add the tests (that are commented out in the R files) as an article perhaps.

Some of the R code in R/ folder contains tests which are commented. Nice to convert these to unit tests.

Nice to see the path diagram for the factor models.

The Rao-Scott adjustment seems natural to use in the context of complex sampling. The problem is it requires a Choleski decomposition of $\boldsymbol\Omega_2$, whose estimate can be not full rank. Could we explore stable/modified Choleski decompositions and see how it affects the $p$-value calculations?

install.packages("Matrix")
library(Matrix)

# Create a non-positive definite matrix
A <- matrix(c(4, 2, 2, 2, 4, 2, 2, 2, 4), nrow = 3, ncol = 3)

# Find the nearest positive definite matrix
nearest_pd <- nearPD(A, corr = FALSE, do2eigen = TRUE, ensureSymmetry = TRUE)

# Perform Cholesky decomposition on the nearest positive definite matrix
L <- chol(nearest_pd$mat)

print(L)

MOJ (2008): Two strategies exist to compute the GOF tests. Choose $W$ such that

1. $X^2 = \mathbf e^\top \boldsymbol \Xi \mathbf e$ is computationally easy to obtain.
2. $X^2= \mathbf e^\top \boldsymbol \Xi \mathbf e$ is asymptotically $\chi^2$.

Our tests that satisfy criteria 1:

• Wald diagonal
• RSS
• Multinomial

Our tests that satisfy criteria 2:

• Wald
• Wald VCF
• Pearson