test to make sure dsem output is consistent
You can install the development version of testingdsem from GitHub with:
# install.packages("devtools")
devtools::install_github("andybeet/testingdsem")
-
See figures below
-
Each figure shows a fitted network as found in the dsem vignette
-
The original model fit is show with 3 other fits where the only difference is that some data have been set to missing (NAs)
-
Each of the three figures show three random assignments of NAs. In each case 25% of the data was assigned missing.
-
The parameters seem to be estimated ok, the algorithm seems to have converged ok, and the parameters that look quite different are statistically significant.
-
Note in particular the paths time->pwage and cprofits->invest
Fit from vignette
library(testingdsem)
plotid <- testJim()
#> 1 regions found.
#> Using 1 threads
#> 1 regions found.
#> Using 1 threads
plotid$porig
Fits of same model with missing data
plotid$p1
plotid$p2
plotid$p3