Fixed some return; calls in the import function.
Minor changes to provide a clean R CMD check
Minor bug fix in import()
Updated the import function so that three indicators are possible.
Fixed minor error in examples.
fSRM now gives an error if users try to compare more than two groups.
Updated CITATION file
Changed order of dyads in the mean-structure plot. Order now is the same as in the variance-decomposition plot.
Prettier output for equalMeans function.
Fixed a bug when variances where constrained to be positive in a model with group comparisons and no mean structures: In this case, only the variances of the first group were constrained. Now it's fixed: both groups are constrained.
Added new data sets: clinical, clinical.wide, two.groups, two.groups.wide
When running a two-group analysis, plots automatically show both groups side-by-side
Fixed a bug where the ordering of labels in the meanstructure-plot was wrong
Removed setZero option from fSRM. Replaced it with the option noNegVar. If this is TRUE, all variances are constrained to be > 0. (Older publications contrained negative variances to be exactly zero. This, however, leads to a distortion of the df). By default, this option is turned off, as negative variances should be recognized as an indicator of model misspecification.
Added rolesEqual option to fSRM: Do roles matter at all? Compare a model with free roles (m1 <- fSRM(..., means=TRUE, rolesEqual = FALSE)) and a model with equal roles (m2 <- fSRM(..., means=TRUE, rolesEqual=TRUE)) using anova(m1$fit, m2$fit). This is a model with no mean difference, the actor variances equal, partner variances equal, relationship variances equal, and the respective reciprocities equal (Thanks to a suggestion of David Kenny).
If missing values are present, by default missing = "fiml" is set, and a warning is issued that this is only valid if the data are missing completely at random (MCAR) or missing at random (MAR). Use missing = "listwise" to exclude families with missing values.
Moved all dependencies to Imports so that no manual installation of packages is necessary for unexperienced users.
By default, one-sided p values and CIs are printed for variances.
Added plotting methods for fSRM objects: plot(x) gives a plot of the relative variances; plot(x, means=TRUE) gives a plot of the mean structure decomposition.
Significance stars are printed in the output & nicer formatting of p values.
Bootstrapping standard errors now properly works (set se="boot" in the fSRM function.)
Out-of-bound estimates of correlations are set to NA in the output.
Re-included the import function, and added the getImport function. These functions can be used to import and transform a data set in wide format.
Added parameter pairwise to the fSRM function (you also have to set means=TRUE that it becomes active). This provides all pairwise comparisons of actor and partner means between roles (e.g., is the actor mean of mothers higher than that of fathers?)
Fixed a bug in print.fSRM (Differences of variances between groups were printed incorrectly)
Included examples in the fSRM help file.
Pretty inst/News.Rd file
Minor changes for passing R CMD check ...
renamed data sets and changed some column names from the output
Now works with missing values. Default behavior: Families with missing values are excluded (listwise deletion). Set fSRM(..., missing="fiml") for ML / FIML estimation
More consistent behavior of parameters means, diff, and group in fSRM. It does now that what would be expected.
mod() function now works with mean structure and delta method if lavaan >= 0.5.16 is installed
added a data set based on Eichelsheim (2011): data(two.groups)
Fixed minor bug in print.fSRM with multiple groups
Now prints a warning if it is attempted to automatically set negative variances to zero in the multiple-group case (Not implemented yet).
Added helper functions which help to import data sets from different formats: 'import()'
Negative variances can be automatically set to zero: setZero = "negative", or set all non-significant variances to zero: setZero = "nonsig". Please note: The purpose of this function is to reproduce published results; usually it is *not* recommended to set non-significant variances to zero! This parameter does not work yet with multiple groups.
Included two data sets, based on Bill Cook's (2000) study: Attachment anxiety in 4-person families (three.person and four.person, see also http://davidakenny.net/kkc/c9/c9.htm), and attachment dependency (two.indicators)
renamed parameter for 'drop': before: "reciprocities", now: "GR"
renamed parameter for fSRM: before: "delta", now: "diff"
minor fixes
Delta method now also compares variances between two groups
Delta method works for 3-person families, and with dropping of factors
equalMeans() now provides a Wald test for equality of relationship effect means
drop does not remove the factor, but rather sets its variance to zero.
Improved output of the mod() function
Mean structure for 3-person family now works correctly
The family effect now is included in the delta method
Replaced 'fe' parameter in fSRM by 'drop': Now the user can drop either the family effect, the actor, the partner effects, or the reciprocities when only three members are present.
New function equalMeans() tests actor and partner means for equality
Implemented delta method: all means of two groups can be compared simultaneously. Set delta=TRUE and group='group' in fSRM, and you're done!
The structure of the raw data is not shown anymore when executing the fSRM function
Fixed some inconsistencies in labeling
Preliminarily dropped analyses with self-ratings (they probably will be added later again)
Removed error correlations type ‘within’: This has been described in the literature, but actually never been used.