pams - Profile Analysis via Multidimensional Scaling
Implements Profile Analysis via Multidimensional Scaling
(PAMS) for the identification of population-level core response
profiles from cross-sectional and longitudinal person-score
data. Each person profile is decomposed into a level component
(the person mean) and a pattern component (ipsatized
subscores). PAMS uses nonmetric multidimensional scaling via
the SMACOF algorithm to identify a small number of core
profiles that represent the central response patterns in a
sample of any size. Bootstrap standard errors and
bias-corrected and accelerated (BCa) confidence intervals for
individual core profile coordinates are estimated, enabling
significance testing of coordinates that is not available in
other profile analysis methods such as cluster profile analysis
or latent profile analysis. Person-level weights, R-squared
values, and partial correlations with core profiles are also
estimated, allowing individual profiles to be interpreted in
terms of the core profile structure. PAMS can be applied to
both cross-sectional data and longitudinal data, where core
trajectory profiles describe how response patterns change over
time. Methods are described in Kim and Kim (2024)
<doi:10.20982/tqmp.20.3.p230>, de Leeuw and Mair (2009)
<doi:10.18637/jss.v031.i03>, and Kruskal (1964)
<doi:10.1007/BF02289565>.
