umxMI: Report modifications which would improve fit.

In umx: Structural Equation Modeling and Twin Modeling in R

Report modifications which would improve fit.

Description

This function uses the mechanical modification-indices approach to detect single paths which, if added or dropped, would improve fit.

Usage

umxMI(
  model = NA,
  matrices = NA,
  full = FALSE,
  numInd = NA,
  typeToShow = "both",
  decreasing = TRUE
)

Arguments

model	An OpenMx::mxModel() for which to report modification indices
matrices	which matrices to test. The default (NA) will test A & S for RAM models
full	If FALSE (default), only the parameter under test can move. If TRUE, all free parameters re-adjust (much slower).
numInd	How many modifications to report. Use -1 for all. Default (NA) will report all over 6.63 (p = .01)
typeToShow	Whether to shown additions or deletions (default = "both")
decreasing	How to sort (default = TRUE, decreasing)

Details

Notes:

1. Default full = FALSE only moves the parameter under test (fast screen). Set full = TRUE to allow all free parameters to re-adjust (slower; closer to a full re-fit for each candidate).

2. Compared to mxMI, this function returns top changes, and also suppresses the run message.

3. Finally, of course: see the requirements for (legitimate) post-hoc modeling in OpenMx::mxMI() You are almost certainly doing better science when testing competing models rather than modifying a model to fit.

References

https://github.com/tbates/umx

See Also

• OpenMx::mxMI()

Other Model Summary and Comparison: umx, umxCompare(), umxEquate(), umxReduce(), umxSetParameters(), umxSummary(), xmu_compare_WLS(), xmu_robust_WLS_fit()

Examples

## Not run: 
require(umx)
data(demoOneFactor)
manifests = names(demoOneFactor)
m1 = umxRAM("One Factor", data = demoOneFactor, type = "cov",
	umxPath("G", to = manifests),
	umxPath(var = manifests),
	umxPath(var = "G", fixedAt = 1)
)
umxMI(m1)
umxMI(m1, full = TRUE)  # slower: allow all parameters to re-adjust

## End(Not run)

umx documentation built on Oct. 11, 2026, 5:07 p.m.