0.3 series

Version 0.3-3
  • Released on CRAN: 19 Januari 2011
  • only a fix for a missing link in the documentation
Version 0.3-1
  • Released on CRAN: 11 May 2010
  • New features/changes for this version (compared to the older program semplus that was never released on CRAN):
    • name of the package has changed to ‘lavaan’
    • ‘ML.N’ option is replaced by a ‘mimic.Mplus’ option
    • if do.fit=FALSE, a full summary (including standard errors) is now available
    • if a correlation matrix is supplied (instead of a covariance matrix), only a (big) warning is now spit out (instead of an error and stopping)
    • model syntax can now be specified as a string literal enclosed in single quotes
    • multiple values are now accepted within pre-multiplication commands when analyzing multiple groups
    • in a multiple group analysis, the sample moments can be provided using a list
    • using NA*x in a formula forces the corresponding parameter to be free
    • a new modifier ‘label’ can now be used to specify custom labels
    • added ‘information’ argument
    • if na.rm=FALSE and estimator=“ML”, full information ML (FIML) is used
  • Known issues for this release:
    • MLM values are different in Mplus (but same as in EQS)
    • SRMR values are sometimes different in Mplus
    • EPCs for equality constraints are wrong
  • Bugs/glitches discovered after the release:
    • If the data frame contained cases where all values were missing, this produced an error
    • In mimic models with a single covariate, the baseline chi-square test statistic was not computed correctly, and the standardized solution gave an error
    • If an observed or latent variance (for example x1 ~~ x1 ) was included in the model syntax without a starting value or a fixed value, the starting value was set to zero
    • lavaan could not handle some non-standard models; for example: latent variables where the indicators are a mixture of latent and observed variables; indicators that are also predictors in a regression
    • lavaan could not handle a model with latent variables and a dependent observed-only variable (for example y ~ f1 + f2 + f3 where f1, f2 and f3 are latent variables, but y is an observed variable)