partwhole: Part-Whole Correlation

Description

Returns the part-whole correlations between an item or the mean of all possible groups of nitems and the composite of the full set of items.

Usage

partwhole(x, nitems = 1, nomiss = 0.8)

Arguments

A matrix or data.frame containing the variables (in columns) thought to form a composite.

nitems

A numeric element indicating the number of items desired for each possible group of items.

nomiss

A numeric between .00 and 1.00 indicating the proportion of scores that must be non-missing for each composite before a score of NA is returned.

Value

A matrix with 2 rows and K columns where K is the number of possible subset combinations. The column names indicate which items (separated by an underline) make up the subset combination. The first row (UnitWgt) is the result using a unit weighted composite for the total set of items and the second row (Component) is the result using principle component scores for the total set of items.

Details

The purpose of this function is to determine which subset of items, when formed into a unit-weighted composite, most strongly correlate with both a unit-weighted and a components weighted composite of the full set of items. For example, if one has an 8 item scale and wants to reduce it to a 4 item scale, it might be interest to know which 4 items can be composited and correlate most highly with the composite from the full set of 8 items. It turns out there are 70 ways to form 4-item composites from 8 total items. This function creates all 70 of those composites and correlates each with both a unit weighted composite from the original 8 items and a components scored (1 principal component) composite of the original 8 items. One can then look at the output to determine which 4-item composite best correlated with the full scale composite.

Examples

Run this code

data(bfi.set)
  # Imagine we want to find the best two-item composite that correlates
  # highest with the full 8 items available to measure extraversion. 
  # Three (of the extraversion) items need to be reverse scored
sBFI6r <- 6 - bfi.set$sBFI6
sBFI21r <- 6 - bfi.set$sBFI21
sBFI31r <- 6 - bfi.set$sBFI31
  # Now put them all into one data.frame
ext.vars <- data.frame(bfi.set$sBFI1, sBFI6r, bfi.set$sBFI11, 
	bfi.set$sBFI16, sBFI21r, bfi.set$sBFI26, sBFI31r, bfi.set$sBFI36)
head(ext.vars) # Looks good
  # Now compute the parwhole correlation for all possible 2-item composites
partwhole(ext.vars, 2)