nFK: Number of Orthogonal Factors Measured by Set

In funfield-lab/fancyr: Fancy Statistics for Correlational (r) Analyses

Number of Orthogonal Factors Measured by Set

Description

For a specified set of variables, estimate the 'effective number of orthogonal factors' measured.

Note that 'measured' is intended to mean that a factor is only measured if it has an R^2 value of 1.00. So statistic provides the sum of R2 for orthogonal dimensions, via PCA.

Usage

nFK(rMat, rotate = "none")

Arguments

rMat	Correlation matrix (with appropriate reliabilities on diagonal) The correlation matrix should have a unique name for each variable, and these variable names should be a single word or string (no spaces). Can check with colnames(rMat)

Details

The correlation matrix should have correct reliability values on the diagonals (ideally, the value will be estimates of the retest values over the same measurement interval as typical for inter-item correlations within the matrix; see Wood, Lowman, Armstrong, & Harms, 2022). If values of 1.0 are used on the diagonal rather than correct reliability estimates, then the resulting nFK estimate will be inflated - often substantially. However, it may still be useful to estimate nFK in this case to provide an 'upper-bound' estimate of nFK.

There is also some code that is a work in progress to try to do this through 'setCor', although it is commented out as it has problems at the moment.

Value

returns: (1) the PCA loadings, (2) the R^2 for each separate factor, and (3) the nFK for the entire item set (sum of R^2)

funfield-lab/fancyr documentation built on Nov. 21, 2023, 2:42 p.m.