NEWS

---

nonnest2 0.5-8 (2024-08-28)

• add support for objects of DiscreteClass from mirt package (credit to Phil Chalmers)
• bug fix and improved check for object calls

nonnest2 0.5-7 (2024-05-06)

• internal change so that nonnest2 is more likely to work with new object classes

nonnest2 0.5-6 (2023-08-13)

• add initial functionality for OpenMx models (credit to Mauricio Garnier-Villarreal)

nonnest2 0.5-4 (2020-06-19)

• fix bugs for lavaan models with missing data, where some rows of the data matrix were fully missing

nonnest2 0.5-3 (2020-03-14)

• simplify llcont() for mlogit models (credit to Yves Croissant)
• fix for AIC/BIC adjustments to test statistics (credit to Thomas Hielscher)
• handle sampling weights for mirt models, throw error for lavaan sampling weights

nonnest2 0.5-2 (2018-09-25)

• new vuongtest() arguments ll1, ll2, score1, score2, vc1, vc2 for sending in custom functions that obtain casewise log-likelihood, scores, and vcovs from fitted models
• add packages to Suggests, fix a failing test

nonnest2 0.5-1 (2018-01-22)

• llcont() for lavaan objects with fixed.x=TRUE is corrected

nonnest2 0.5 (2017-12-15)

• include llcont() method for mirt objects (credit to Lennart Schneider) and find mirt's estfun.AllModelClass()
• lavaan changes: allow for inclusion of completely missing cases and stop if missing = "listwise" or "pairwise"
• return error for lavaan models with fixed.x=TRUE (incorrect results)
• fix scaling issues in vcov associated with lm/glm/nls

nonnest2 0.4

• Update handling of missing data to conform to lavaan 0.5-21.
• Throw error in case lavaan model with robust SEs is sent in.

nonnest2 0.3 (2016-01-23)

• Added "Getting started" vignette.
• Changes to conform to lavaan 0.5-20's handling of equality constraints.
• Fixed issue where print methods sometimes wrapped over lines.
• Fixed bug in llcont() for lavaan cases missing all data except one variable.

nonnest2 0.2 (2014-12-04)

• Added argument "nested" to vuongtest(), allowing the user to specify whether or not candidate models are nested. This impacts the limiting distribution of the Vuong LRT.
• Added argument "adj" to vuongtest(), allowing the user to obtain test statistics with AIC or BIC adjustments.
• print methods no longer display the full call of each model (just the first line).
• BIC intervals added for objects of class hurdle, zeroinfl, mlogit (were missing because they didn't have an nobs method).