fSRM R Documentation

## News for package 'fSRM'

### Version 0.6.5

#### Changes

• Fixed some return; calls in the import function.

### Version 0.6.4

#### Changes

• Minor changes to provide a clean R CMD check

### Version 0.6.3

#### Changes

• Minor bug fix in import()

### Version 0.6.2

#### Changes

• 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

### Version 0.6.1

#### Changes

• Changed order of dyads in the mean-structure plot. Order now is the same as in the variance-decomposition plot.

### Version 0.6

#### Changes

• 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

### Version 0.5

#### Changes

• 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.

### Version 0.3.5

#### Changes

• 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.

### Version 0.3.3

#### Changes

• 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?)

### Version 0.3.2

#### Changes

• Fixed a bug in print.fSRM (Differences of variances between groups were printed incorrectly)

• Included examples in the fSRM help file.

### Version 0.3.1

#### Changes

• Pretty inst/News.Rd file

• Minor changes for passing R CMD check ...

### Version 0.3

#### New Features

• 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.

### Version 0.2.6

#### New Features

• 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

### Version 0.2.2

#### New Features

• 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

#### New Features

• 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

### Version 0.2.1

#### New Features

• 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!

#### Changes

• 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.