I just submitted my package update (version 1.3) to CRAN. The download is already available (currently source, binaries follow). While the last two updates included new functions for table outputs (see here and here for details on these functions), the current update mostly provides small helper functions. The focus of this update was to improve existing functions and make their handling easier and more comfortable.

Automatic label detection

One major feature is that many functions now automatically detect variables and value labels, if possible. For instance, if you have imported an SPSS dataset (e.g. with the function sji.SPSS), value labels are automatically attached to all variables of the data frame. With the autoAttachVarLabels parameter set to TRUE, even variable labels will be attached to the data frame after importing the SPSS data. These labels are automatically detected by most functions of the package now. But this does not only apply to importet SPSS-data. If you have factors with specified factor levels, these will also automatically be used as value labels. Furthermore, you can manually attach value and variable labels using the new function sji.setVariableLabels and sji.setValueLabels.

But what are the exactly the benefits of this new feature? Let me give an example. To plot a proportional table with axis and legend labels, prior to sjPlot 1.3 you needed following code:

efc.val <- sji.getValueLabels(efc)
efc.var <- sji.getVariableLabels(efc)

Since version 1.3, you only need to write:

sjp.xtab(efc$e16sex, efc$e42dep)

Reliability check for index scores

One new table output function included in this update is sjt.itemanalysis, which helps performing an item analysis on a scale or data frame if you want to develop index scores.

Let’s say you have several items and you want to compute a principal component analysis in order to identify different components that can be composed to an index score. In such cases, you might want to perform reliability and item discrimination tests. This is shown in the following example, which performs a PCA on the COPE-Index-scale, followed by a reliability and item analysis of each extracted „score“:

# recveive first item of COPE-index scale
start <- which(colnames(efc)=="c82cop1")
# recveive last item of COPE-index scale
end <- which(colnames(efc)=="c90cop9")
# create data frame of cope-index-items
df <- as.data.frame(efc[,c(start:end)])
colnames(df) <- sji.getVariableLabels(efc)[c(start:end)]
# compute PCA on cope index and return
# "group classifications" of factors
factor.groups <- sjt.pca(df, no.output=TRUE)$factor.index
# perform item analysis
sjt.itemanalysis(df, factor.groups)

The result is following table, where two components have been extracted via the PCA, and the variables belonging each component are treated as one „index score“ (note that you don’t need to define groups, you can also treat a data frame as one single „index“):

The output of the computed PCA was suppressed by no.output=TRUE. To better understand the above figure, take a look at the PCA results, where two components have been extracted:

Beside that, many functions – especially the table output functions – got new parameters to change the appearance of the output (amount of digits, including NA’s, additional information in tables etc.). Refer to the package news to get a complete overview of what was changed since the last version.

The latest developer build can be found on github.