dmacs (version 0.1.0)

dmacs_summary: Summary of measurement nonequivalence effects

Description

dmacs_summary returns a summary of measurement non-equivalence effects given lists of parameters.

Usage

dmacs_summary(LambdaList, ThreshList, MeanList, VarList, SDList,
  Groups = NULL, RefGroup = 1, categorical = FALSE, ...)

Arguments

LambdaList

is a list, indexed by groups, of factor loading matrices (dataframes are allowed).

ThreshList

is a list, indexed by groups, of vectors of indicator intercepts (for continuous indicators) or lists, indexed by items, of vectors of thresholds (for categorical indicators). For categorical indicators, do not provide a matrix of thresholds for each group.

MeanList

is a list, indexed by groups, of vectors of factor means. For unidimensional models, this is simply a list of factor means.

VarList

is a list, indexed by groups, of vectors of factor variances. For unidimensional models, this is simply a list of factor variances.

SDList

is a list, indexed by groups, of vectors of indicator observed standard deviations used as the denominator of the dmacs effect size. This will usually either be pooled standard deviations or the standard deviation of the reference group. Each group, including the reference group, must be included in SDList (although the standard deviations for the reference group are ignored).

Groups

is a vector of group names. If no value is provided, dmacs_summary will try to use names(LambdaList); if LambdaList has no names, then the groups will be numbered.

RefGroup

can be the name of the reference group (as a string), or the index of the reference group (as a number). RefGroup defaults to the first group if no value is provided. It is strongly recommended to provide the reference group as a string, since group names in data are often ordered by their appearance in the data, not alphabetically.

categorical

is a Boolean variable declaring whether the variables in the model are ordered categorical. Models in which some variables are categorical and others are continuous are not supported. If no value is provided, categorical defaults to FALSE, although if multiple thresholds are provided for an item, categorical will be forced to TRUE. A graded response model with probit link (e.g., DWLS in lavaan or WLSMV in Mplus) is used for categorical variables. If you desire for other categorical models (e.g., IRT parameterization) to be supported, e-mail the maintainer.

...

other parameters to be used in functions that dmacs_summary calls, most likely stepsize for the item_dmacs and delta_mean_item functions.

Value

A list, indexed by groups, of lists of measurement nonequivalence effects from Nye and Drasgow (2011), including dmacs, expected bias in the mean score by item, expected bias in the mean total score, and expected bias in the variance of the total score. Expected bias in the variance of the total score is only supplied for unidimensional models with linear indicators (i.e., not categorical) in the current version of this package.

Details

dmacs_summary is called by lavaan_dmacs and mplus_dmacs, which are the only functions in this package intended for casual users

References

Nye, C. & Drasgow, F. (2011). Effect size indices for analyses of measurement equivalence: Understanding the practical importance of differences between groups. Journal of Applied Psychology, 96(5), 966-980.

Examples

# NOT RUN {
LambdaList <- list(Group1 <- matrix(c(1.00, 0.74,  1.14, 0.92), ncol = 1),
                   Group2 <- matrix(c(1.00, 0.76,  1.31, 0.98), ncol = 1))
ThreshList <- list(Group1 <- c(0.00, 1.28, -0.82, 0.44),
                   Group2 <- c(0.00, 0.65, -0.77, 0.47))
MeanList   <- list(Group1 <- 0.21,
                   Group2 <- 0.19)
VarList    <- list(Group1 <- 1.76,
                   Group2 <- 1.34)
SDList     <- list(Group1 <- c(2.12, 1.85,  1.12, 3.61),
                   Group2 <- c(NA, NA, NA, NA))
Groups <- c("Group1", "Group2")
RefGroup <- "Group2"
dmacs_summary(LambdaList, ThreshList, MeanList, VarList, SDList,
              Groups, RefGroup)

# }