phatd: Reliability-adjusted Correlations Over a Measurement Interval...

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

Reliability-adjusted Correlations Over a Measurement Interval (d)

Description

Estimate the degree to which correlations between variables over measurement interval (d; generally a period of time, or number of items apart) are smaller than the retest correlations of the variables over that interval. This is estimated via an understanding of Spearman's adjustment("correction") for measurement unreliability.

For instance: phat_xy(d) = .80 indicates that the correlation between x and y over measurement interval t is 80% of the retest correlations of X and Y (i.e., their correlations with themselves) over the same measurement interval.

Importantly: x1 and x2 should be the same set of variables, given in the same order, but rated at different times. For instance, x1 and x2 may be the first and second administrations of the items in some personality inventory a year apart (or: a month, a day, etc).

Usage

phat.t(x1, x2)
phat.t(x1, x2, posdef = "Higham")
phat.t(x1, x2, adjust = F, posdef = "Higham") #don't adjust for retest reliability

Arguments

x1	first matrix to be used to estimate correlations
x2	second matrix to be used (all variables should be the same as x1, and in the same order!!)
adjust	adjust correlations for retest reliability? Defaults to TRUE
posdef	adjust correlations to make the matrix positive definite? ("Bock", "Higham", and default = FALSE)

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